Why Ethnic Groups Rebel

Home , Baluch (Uttar Pradesh), Dayton agreement, Florin Krasniqi, Franjo Kluz, Kosovo Protection Corps, Likoshan

Why Ethnic Groups Rebel:

Intra-ethnic Division, Dynamic Grievances, State Repression and Escalation

Keiichi KUBO

Thesis submitted for the degree of PhD in Political Science

London School of Economics and Political Science Department of Government

London, 2007

1 \ UMI Number: U501706

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion.

Dissertation Publishing

UMI U501706 Published by ProQuest LLC 2014. Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 {■ 6? 6 4 - &A, \im *4-

:btbH gqutnOyindl3 vri W

t^anK79hO oimftn/fl nobr/id aindio-B'rtnl

iio h d im 3 bne noi0«3iqs^l

OflUi' ifbibX

to osigoh 3ffj lot bstfjimdu* zi^riT ooivjio?. LeomloH ni Gricl

oonob8 IsdrliloH brm .goimonooH r!'o loorb2 riobnoJ JnornrnovoO 'to tnomttBq >(I

U)()i nobno.l Library

jjrii& rrnr.'i *: '■” • Declaration

I hereby certify that the work presented in the thesis is solely my own work other than where I have clearly indicated that it is the work of others.

Tokyo, 30 November, 2007

KeiichiKUBO

2 Abstract

Why do ethnic groups rebel against the state? While there have been various efforts to answer this question, existing explanations tend to be static and based on the unitary actor assumption. This thesis attempts to make a step forward by relaxing the unitary actor assumption and by introducing dynamic elements into the explanatory framework. In order to do so, it takes a three-step approach, examining (1) the initial intensity of rebellion, (2) the onset of rebellion, and (3) the escalation. As for the methodology, this thesis adopts a mixed-methods approach: for each step, it conducts both large-N analyses and qualitative analyses of ex-Yugoslav countries. This thesis firstly examines the determinants of the initial intensity of rebellion. It presents two ideal-types, “decisions from above” made by ethnic leaders and “decisions from below” made by non-leaders, and will argue that the initial intensity is more likely to be higher if ethnic leaders organise the rebellion, because they can mobilise more resources. Secondly, this thesis examines the causes of the onset of rebellion. It will argue that structural conditions alone do not suffice to explain it, and will argue for the importance of dynamic grievances. In other words, even when structural conditions do not change, grievances will increase over time as the duration of peaceful protests gets longer, because people will be increasingly frustrated by the failure to achieve their goals by peaceful means. Finally, this thesis examines the dynamics of escalation. It will be argued that the low-intensity rebellion is likely to provoke state repression, and that the repressive measures taken by the state in turn are likely to cause the escalation of rebellion. The concluding chapter explores the theoretical and methodological implications for the study of ethnic conflict as well as the policy implications for conflict prevention.

3 Contents

Introduction 10

Part I: Theoretical and Methodological Premises

Chapter 1 A Three-Step Approach for the Analysis of Ethnic Rebellion Introduction 18 1. Definition of the Ethnic Rebellion 18 2. Four Problems: Critical Review of the Present Theoretical Literature 23 (1) Unitary Actor Assumption 24 (2) A Controversy - Do Grievances Matter? 27 (3) Lack of Dynamics 31 (4) Lack of Attention to the Escalation of Violence 33 3. Three-Step Approach for the Analysis of Ethnic Rebellion 36 (1) Initial Level of Rebellion 36 (2) Onset of the Initial Rebellion 38 (3) Escalation of the Rebellion 38 Conclusion 39

Chapter 2 Methodological Premises: Large-N analysis and the ex-Yugoslav Case Study Introduction 41 1. Methodology: Why Mixed Methods? 41 2. The MAR Dataset 44 3. The ex-Yugoslav Cases: Croatia, Bosnia, Serbia and Macedonia 49 4. Structure of the Thesis: Connection between Theory, Large-N Analyses and Case Studies 56 Conclusion 58

4 Part II: Initial Level o f Rebellion

Chapter 3 Intra-ethnic Division, Decision-Makers and the Rebellion: Quantitative Analysis of the Initial Level of Rebellion Introduction 60 1. Two Ideal Types: Decisions From Above and Decisions From Below 61 2. Drawing Testable Hypotheses 68 3. Testing Hypotheses: Large-N Analyses 75 (1) Bi-variate analysis 76 (2) Multi-variate analysis 81 Conclusion 88

Chapter 4 Decisions from Above and from Below: Comparative Analysis of the Serb and Albanian Rebellions Introduction 89 1. Serbs in Croatia 91 (1) Involvement of Politicians and Police Forces in the Rebellion 91 (2) Involvement of the Regular Military Force in the Rebellion 95 (3) High Level of the Initial Intensity of Rebellion 97 2. Serbs in Bosnia 98 (1) Involvement of Politicians in the Rebellion 98 (2) Involvement of the Regular Military Force in the Rebellion 101 (3) High Level of the Initial Intensity of Rebellion 102 3. Albanians in Kosovo 104 (1) Path Chosen by the Albanian Elite in Kosovo: Peaceful Resistance 104 (2) Decision from Below: the KLA and Its Limited Resources 107 (3) Low Level of the Initial Intensity of Rebellion 109 4. Albanians in Macedonia 110 (1) Path Chosen by the Albanian Elite: Moderation and Participation 110 (2) Decision from Below: the NLA and Its Limited Resources 114 (3) Low Level of the Initial Intensity of Rebellion 115 Conclusion 116

5 Part III: Onset o f Initial Rebellion

Chapter 5 Etiology of Ethnic Rebellion: Quantitative Analysis of the Effect of Dynamic Grievances Introduction 118 1. Theoretical Model: Dynamic Grievances and Structural Conditions 118 (1) Structural Conditions 121 (2) Dynamic Grievances: Importance of the Time Dimension 126 (3) The Effect of the Type of Decision-Makers 129 2. Large-N Analysis 131 (1) Research Design and Operationalisation 132 (2) Empirical Results 138 Conclusion 146

Chapter 6 Dynamic Grievances and Structural Conditions: Qualitative Analysis of the Ex-Yugoslav Countries Introduction 149 1. Albanians in Kosovo 150 (1) Structural Conditions 150 (2) Dynamic Grievances and the Availability of Arms 153 2. Albanians in Macedonia 159 (1) Structural Conditions 159 (2) Dynamic Grievances and the Availability of Arms 162 3. Serbs in Croatia 167 (1) Motivating Conditions 167 (2) Enabling Conditions 169 4. Serbs in Bosnia 172 (1) Motivating Conditions 173 (2) Enabling Conditions 175 Conclusion 177

Part IV: Dynamics of Rebel Escalations

Chapter 7 State Repression as an Intervening Variable: Quantitative Analysis of Rebel Escalations Introduction 180

6 1. Theoretical Exploration 181 (1) Dynamics between the Ethnic Group and the State 181 (2) Structural Factors for State Repression 184 (3) Structural Factors for the Escalation of Rebellion 185 2. Large-N Analysis 187 (1) Research Design 187 (2) Empirical Results 195 Conclusion 203

Chapter 8 Repression or Conciliation? Comparative Case Study of Serbia and Macedonia Introduction 205 1. Case of Serbia: Military Repression and the Escalation 205 (1) State Reactions to the Initial Rebellion 205 (2) Effect of Repression: Expansion of the KLA 208 (3) Second-Phase Operations: Serbian Summer Offensive 212 (4) From the Holbrooke Agreement to Rambouillet 215 2. Case of Macedonia: Peace Process and Prevention of Escalation 217 (1) Trajkovski’s Political Process for the Solution of Conflict 218 (2) Negotiations at Ohrid 220 (3) Importance of Trajkovski’s Political Process 223 Conclusion 225

Chapter 9 Explaining State Reactions: Path Dependency and External Pressures in Serbia and Macedonia Introduction 226 1. Explanatory Factors: Path Dependency and External Pressures 227 2. Case of Serbia 229 (1) Path-dependency and the Context of Domestic Politics 229 (2) Ambiguous Signals from External Actors 235 3. Case of Macedonia 239 (1) Path Dependency: Election with the Support of Albanians 239 (2) External Pressures from the International Community 243 Conclusion 248

7 Conclusion 1. Key Findings of the Thesis 251 2. Implications of the Case Selection 253 3. Theoretical and Methodological Implications 257 (1) Unitary Actor Assumption 257 (2) Conceptualisation and Operationalisation of Civil War 258 4. Policy Implications 260 5. Tasks for Future Research: Additional Implications and Theoretical Extensions 262

Reference 265

Appendices 283

Tables and Figures

Tables 1.1 Initial Level, Onset and Escalation 37 3.1 Type of Decision-makers and the Initial Level of Rebellion 66 3.2 Status of the Ethnic Group within the State and the Initial Level of Rebellion 67 3.3 Type of decision-makers, the Status of the Ethnic Group and the Initial level of the rebellion 68 3.4 Initial Level of Rebellion and Involvement of Politicians, 177 ethnic groups from 1945 to 2000 78 3.5 Initial Level of Rebellion and Involvement of Police or Army, 177 ethnic groups from 1945 to 2000 80 3.6 Ordered Logit Analysis of the Initial Level of Rebellion, 177 ethnic groups 85 5.1 Random-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: REBONSET) 139 5.2 Random-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: REBONSET2) 140 5.3 Random-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: LOWONSET) 143

8 5,4 Random-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: HIGHONSET) 145 6.1 GDP Real Growth Rates in Macedonia, 1991-2000 162 7.1 Ordered Logit Analysis of the Determinants of Military Repression, 1996-2000 (Dependent Variable: REP20) 196 7.2 Random-Effects Logit Analysis of Determinants of the Escalation of Rebellion, 1996-2000 198 7.3 Ordered Logit Analysis of the Determinants of the Level of Rebellion, 1996-2000 (Dependent Variable: REB) 199 7.4 Cross-Tabulation of Rebellion Index and the Level of Repression (Number of Instances, 1996-2000) 201 7.5 Previous Involvement in the Rebel Activities by the Targeted Ethnic Group, 1996-2000 202 9.1 Results of the Presidential Elections in Macedonia, 1999 240

Figures 5.1 Effect of the Discounting Factor 128 7.1 Dynamic Relationship between Rebellion and Repression 183

9 Introduction

Since the end of the Cold War, increased attention has been given to ethnic conflict and civil war. While such conflicts are not new,1 tragic events in such places as ex-Yugoslavia and Rwanda surprised many people in the

West who were in the euphoria of triumph over its rival, and became a favoured topic for academic research and media coverage. The ongoing civil strife in Iraq shows that such conflicts still constitute one of the main threats to peace and security in the world. To examine the causes of ethnic conflict and civil war certainly remains an important task for researchers.

Asking why ethnic groups rebel is, therefore, important. This is a classic question, however, and merely asking it is hardly innovative. Gurr

(1970) asked why men rebel already in the 1970s. Horowitz (1985) analysed ethnic groups in conflict in the 1980s. Furthermore, since the end of the Cold

War, there has been a remarkable development in the study of ethnic conflict and civil war, and various authors have attempted to build new explanatory models and to identify the causes of conflict based on wide ranging large-N analyses.

The originality of the present thesis does not lie in the question it asks, but in the manner in which it answers the question. As discussed in Chapter 1, existing explanations tend to be static and based on the unitary actor assumption. This thesis attempts to make a step forward by relaxing the unitary actor assumption and introducing dynamic elements into the explanatory framework. The unitary actor assumption is relaxed by incorporating the

1 For example, the number of ongoing civil wars reached 25 already by the mid 1980s, the same number as in 1999. Between 1945 and 1999, “conservative estimate” o f the total dead as a direct result of these wars reached 16.2 million, five times larger than that of inter-state wars in the same period. See Fearon & Laitin 2003: 75-77.

10 intra-ethnic divisions and the presence of different types of decision-makers

that exist within ethnic groups. Dynamic elements enter the explanatory

framework (1) by examining the effect of “dynamic grievances,” i.e. a type of

grievances that increases over time even when structural conditions remain the

same, and (2) by examining an interaction between the ethnic group and the

state. Needless to say, this thesis only marks a beginning, rather than an end, of the efforts that will be needed to develop a more dynamic theoretical

framework to explain the occurrence of ethnic conflict and civil war. There is much left to be done, and such efforts must be continued further in future research.

Methodologically, this thesis adopted a “mixed methods” approach that is becoming more and more popular in political science. In short, this approach calls for an interaction between the large-N (quantitative) analysis

and the small-N (qualitative) analysis. In the course of the research for the present thesis, there was a constant tension between the large-N analyses and the qualitative case studies, though such a tension may not be so visible in the present thesis as a final product of the interaction. This tension was not easy to overcome, and the author went back and forth between the two kinds of analyses to find a middle ground between them. But this tension between the large-N analyses and the case studies was useful for the refinement of the theoretical framework.

As for the case study, the ex-Yugoslav region is extensively analysed in the present thesis, based on the results of the fieldwork in the ex-Yugoslav region from August 2005 to October 2006. Besides the fact that I have been interested in this region and have some first-hand experiences of it since 2000, there are many reasons for the case selection. Firstly, the intra-regional comparison often allows us to control for many macro-variables in the

11 comparative qualitative analysis and it is thus a common choice for the “most

similar systems design” (Przeworski & Teune 1970). In addition, the

systematic intra-regional comparison has been relatively rare when it comes to the analysis of violent conflict in ex-Yugoslavia, which makes it possible to make some contribution to the study of ex-Yugoslav conflict. Finally, the original theoretical arguments in this thesis are all inspired primarily by the analysis of the ex-Yugoslav cases. The analysis of the ex-Yugoslav cases enabled me to identify some factors that are not considered in the existing literature and yet were important in the ex-Yugoslav cases. This then became a starting point for the theory-building and theory-testing by the large-N analyses. This case selection is not meant to choose a representative sample.

Note that the large-N analyses of this thesis will cover more than 200 ethnic groups from all over the world and a long period from 1945 up to 2000.

Compared to this wide variety of area and period, the qualitative case study only analyses the cases from one specific region (ex-Yugoslavia) that occurred in the same period (in the 1990s). Therefore, one may question whether theoretical arguments derived from the comparative analysis of these cases are really generalisable: they may well be biased by some factors specific to the area or the period analysed in the case study, and the four cases analysed here may simply be outliers or anomalies in the sample. The large-N analysis serves to check if these questions or doubts can be dismissed. If the hypotheses derived from the case study are supported by the large-N analyses using a larger and more representative sample, this means that these hypotheses are not necessarily biased by the peculiarities of the ex-Yugoslav cases (for the merits of the large-N study, see also Chapter 2).

However, the large-N analysis is not a panacea for the problem of

12 generalisability. There may be some factors that cannot be included in the large-N analyses for technical or other reasons, and yet these factors may play an important role in the ex-Yugoslav cases. The large-N analyses, then, cannot estimate the effect of these factors, and consequently, cannot control for these factors in order to assess the explanatory power of the variables of one’s interest. Indeed, some factors that played a considerable role in explaining the rebellion in the ex-Yugoslav region are not included in the large-N analyses of this thesis, mainly because it is difficult to quantify them in a methodologically convincing manner. These factors imply either the importance of the peculiarities of the ex-Yugoslav cases or the limits of the large-N analysis. The concluding chapter will elaborate on these implications. For the case study, the present thesis uses a variety of materials, such as newspaper articles and other media reports in local languages and English, interviews with local actors and experts, the testimony at the International

Criminal Tribunal for the former Yugoslavia (ICTY), official documents published in the ex-Yugoslav countries, as well as secondary sources (books and journal articles both in local languages and English). As for the media reports, the testimony at the ICTY and the interviews, full references are given in the footnotes. As for the secondary sources, the author’s name and year of publication are given either in the text or in the footnotes, with page numbers after colon if necessary, and their full references are given in the reference list at the end of the thesis. As for the materials taken from websites, the address of the website is given either in the footnotes or in the reference list. All materials in the Cyrillic alphabet are transformed into the Latin alphabet in the footnotes and the reference list for the sake of simplicity.

As is well known, even the name of places is a subject of contention in the ex-Yugoslav region, especially between Serbs and Albanians. For example,

13 Albanians use the name “Kosova” while Serbs use the name “Kosovo and

Metohija” or “Kosmet.” The present thesis uses the name “Kosovo” because it is common practice in English-speaking countries. All towns and villages in

Kosovo have two names, one in Albanian and the other in Serbian. Generally speaking, in the present thesis, places with Albanian majority population are called in Albanian (e.g. Prishtina rather than Pristina), while those with Serb majority population are called in Serbian. These rules, however, are not applied to the quotes, in which expressions used by the speakers or writers are kept without changes.

This thesis consists of four main parts, followed by a concluding chapter. Part I presents the theoretical and methodological premises of the thesis. Chapter 1 presents the theoretical premises, defines ‘ethnic rebellion,’ undertakes a brief but critical review of the existing literature, and finally outlines a three-step approach for the analysis of ethnic rebellion which advocates a step-by-step analysis of three aspects of rebellion, namely (1) the initial level of rebellion, (2) the onset of initial rebellion and (3) the escalation of rebellion. Chapter 2 addresses the methodological issues, discussing the merits of a “mixed-methods” approach and making a brief description of the key dataset used in the large-N analyses as well as four ex-Yugoslav cases examined in the qualitative analyses of the thesis.

Part II, III and IV constitute the main body of the thesis and correspond to the three steps advocated in Chapter 1. Part II analyses the initial level of rebellion. Chapter 3 will present two ideal types, rebellion from above and from below: the former is caused by the decisions to take up arms by the powerful elites, while the latter is caused by decisions of the non-elites. This chapter will argue that the initial level of rebellion is more likely to be higher if ethnic leaders or elites (such as politicians who hold public office, military and

14 police officials) are involved in the organisation of rebellion. The key causal factor that explains the relationship between the type of decision-makers and the level of initial intensity is the amount of resources. Following the resource mobilisation theory, this thesis argues that the amount of resources available to the organizers of rebellion determines the magnitude of initial rebellion. This chapter then draws some testable hypotheses, based on the findings of the comparative case-study of Chapter 4, and conducts large-N analyses to test these hypotheses. Chapter 4 conducts a comparative analysis of Serb and

Albanian rebellions in the ex-Yugoslav region and will show that the involvement of politicians, police and army in the rebellion led to the high level of the initial intensity of rebellion in Croatia and Bosnia, while Albanian rebellions were “from below,” organised by those who did not have a large amount of resources, and consequently the initial level of rebellion was low in

Kosovo and Macedonia.

Part III will examine the onset of initial rebellion. Chapter 5 will argue that structural conditions emphasised in the existing literature may not suffice to explain the onset of initial rebellion and argue for the importance of

“dynamic grievances.” In other words, when members of the ethnic group keep making peaceful protests for a long time, people may get increasingly frustrated by the fact that they cannot achieve their goals by peaceful means.

As a result, some people decide to choose a violent option as an alternative to change the status quo. This chapter thus hypothesise that a rebellion is more likely to occur as the duration of peaceful protest by members of the ethnic group gets longer. It will then conduct a series of large-N analyses to test this hypothesis, controlling for a number of structural conditions. Chapter 6 conducts a case study of the four rebellions in the ex-Yugoslav region and will show that “dynamic grievances” explain the timing of the onset of rebellion in

15 Kosovo and Macedonia, while other structural conditions, such as the newness of the state, better explain the early onset of rebellion in Croatia and Bosnia.

Part IV will examine the escalation of rebellion, focusing on an interaction between the ethnic group and the state. Chapter 7 will argue that this interaction becomes important when one analyses the escalation of rebellion after the onset of initial rebellion. It will hypothesise that the state authorities tend to take repressive measures against the rebels when the low-intensity rebellion occurs, but these repressive measures in turn tend to cause further grievances and incite an escalation of rebellion. The repressive measures are thus often counterproductive, as many authors have pointed out in analysing the process of ethnic conflict. To test these hypotheses, this chapter will conduct some large-N analyses, though they are only preliminary because of the limitations of the MAR dataset. Chapter 8 will compare the state reactions to the initial rebellion in Serbia and Macedonia and will show that the

Serbian authorities took repressive measures against the rebels, which led to the escalation of rebellion, while the Macedonian authorities took a more conciliatory stance, which led to the de-escalation of rebellion. Chapter 9 will then show that this difference between Serbia and Macedonia is caused primarily by two factors, namely the path dependency of policy position taken by key politicians and the nature of the external pressure. The policy position taken by Slobodan Milosevic in the 1980s made it impossible for him to take a conciliatory stance towards Albanian rebels in the 1990s, even if he wanted to.

Furthermore, the external actors, especially the US, sent a wrong signal to

Milosevic by condemning the Kosovo Liberation Army as “terrorists,” that was interpreted by the Serbian authorities as “green light” for repression. In the case of Macedonia, Boris Trajkovski, the then President of Macedonia, was more moderate towards Albanians since his inauguration and thus it was

16 natural for him to take a conciliatory stance, and the external actors put a coherent and clear pressure to find a political solution to the conflict.

As a brief overview above shows, the first chapter in each part will present the theoretical arguments and conduct a large-N analysis to establish some general findings. Then, the next chapters) will conduct an in-depth comparative case-study of the ex-Yugoslav cases on the same subject. The structure of the main body of the thesis will be discussed further in Chapter 2, because it is related to the methodological approach of the thesis. The final chapter concludes with a summary of the key findings and a discussion of the theoretical and methodological implications for the study of ethnic conflict as well as the policy implications for conflict prevention.

17 Part I: Theoretical and Methodological Premises

Chapter 1 A Three-Step Approach for the Analysis of Ethnic Rebellion

Introduction

This chapter presents the theoretical premises of the thesis. Firstly, it presents a working definition of ethnic rebellion. It will then critically review theoretical works on political violence and ethnic conflict, identifying four major problems in the existing literature. Finally, this chapter will present a three-step approach for the analysis of ethnic rebellion.

1. Definition of the Ethnic Rebellion

In attempting to analyse the causes and dynamics of ethnic rebellion, the first task is to define ethnic rebellion. To “define the domain” is particularly important in the study of ethnic conflict, because an extremely vast and diverse phenomenon can be studied under the title of ethnic conflict. Even when one focuses solely on the violence in ethnic conflict, one still finds such diverse themes as pogroms, genocides, riots, xenophobic and anti-immigrant violence, state violence, and so on.1 While these different types of inter-ethnic violence may well be related to each other in some way (and a systematic inquiry into the relationship between different types of inter-ethnic violence has not been

1 See, for example, following works, Klier & Lambroza 1992; Dobkowski & Wallimann 1992; Grimsahw 1969; Bjorgo & Witte 1993; Van Den Berghe 1990; Horowitz 2001. For an excellent review of the works on ethnic and nationalist violence, see Brubaker & Laitin 1998.

18 conducted so far), it would probably turn out to be fruitless if one attempts to explain such a wide range of phenomena by a single theory.2 A definition of ethnic rebellion demarcates the scope of the analysis and helps clarify what will be explained and what will not be explained in the present thesis.

Ethnic rebellion is defined here as violent actions taken by ethnic organisations against the state authorities. This definition includes four key elements, namely (1) the violence, (2) the presence of some organisation behind the violence, (3) the ethnic nature of the organisation, and (4) the state authorities as a target of violence. Let us briefly explore each element of the definition below.

Firstly, ethnic rebellion must involve some violence, whether its intensity is high or low. This is one of the key differences between ethnic rebellion and ethnic conflict. A term ethnic conflict often implies violence, but not necessarily. When ethnic conflict is analysed, such analysis normally examines historical and political backgrounds of inter-ethnic antagonisms and tensions before they get violent, as well as the occurrence of violence. A term ethnic conflict, therefore, implies much wider context than rebellion, which is a more specific, violent outcome that occurs in the broader context of ethnic conflict. This distinction is important, as Varshney (2001: 365-366) argues:

“[o]n the whole, the existing literature has failed to distinguish between ethnic violence and ethnic conflict. Such conflation is unhelpful. In any ethnically plural society that allows free expression of political demands, some ethnic conflict is more or less inevitable, but it may not necessarily lead to violence.”

2 For example, Brown argued as follows: “it is important to recognize that there are many different types of internal conflict, each caused by different things. The challenge for scholars is to identify these different types of conflicts and the different sets o f factors that bring them about. The search for a single factor or set of factors that explains everything is comparable to the search for the Holy Grail - noble, but futile.” See Brown 1997: 4. Brubaker and Laitin also argued that different types of ethnic violence seem to involve “sharply opposed mechanisms and dynamics ” See Brubaker & Laitin 1998: 446.

19 Secondly, ethnic rebellion must involve some organisation. In other words, violent actions must be taken by some members that consciously and actively participate in the organisation, whether it is small or large, in order to be called a rebellion. There is a difference between rebellion and riot, even though both involve violence. A riot can occur due to the spontaneous actions by mobs even when there is no organisation behind it. A rebellion, on the other hand, cannot occur if there is no organisation. Of course, a distinction between the two may not always be easy to make. On the one hand, Brass (1997) pointed out that there is often an involvement of the organised gangs, or what he calls “institutionalized riot system,” behind the large-scale riots while apparently they occur spontaneously. On the other hand, membership of the organisation and the distinction between fighters and civilians is not always clear when it comes to guerrilla organisations.3 However, a significant difference remains between the two. In the case of a riot, a large number of citizens can participate in it easily without consciously choosing to be a member of the organisation that is behind the violence.4 In the case of a rebellion, those who take up arms will not do so without actively choosing to be members of the organisation (even if they participate in the rebellion only part-time).

The third element is the ethnic nature of the organisation. More precisely, the recruitment of members of the organisation must be done primarily according to ethnic lines. This is the key difference between ethnic

3 For example, this lack o f clear distinction between the civil and the military is one of the key aspects o f what Kaldor calls “new war” compared to the “old war” (i.e. international warfare). See Kaldor 2001. 4 Brubaker and Laitin argued, for example, as follows: “an ethnic riot typically involves at one level deliberate manipulation and organisation by a small number o f instigators but also, at other levels, turbulent currents o f crowd behavior governed by powerful emotions and compelling collective representations requiring social psychological and cultural modes o f analysis.” See Brubaker & Laitin 1998: 446.

20 rebellion and rebellion made by, for example, communist groups that recruit their members along class and ideological lines. This element naturally raises one fundamental question: what, then, is ethnicity ? This question could lead to a host of theoretical and empirical literature on ethnicity itself, particularly on the origins of ethnicity, where various schools, such as “primordialist”

“instrumentalist” and “constructivist” approaches, provide sharply opposing explanations.5 An extensive review of these approaches is not presented here.

For present purposes, it is assumed that the ethnic identities and differences already exist in a country when the ethnic rebellion occurs, without asking where these ethnic identities come from. As for the definition of ethnicity, this thesis follows the “inclusive concept” of ethnicity proposed by Horowitz (1985: 17-18), which defines ethnicity “by ascriptive differences, whether the indicum is color, appearance, language, religion, some other indicator of common origin, or some combination thereof.”

Fourthly, a target of violent actions must be the state authorities, such as politicians who hold public offices, government institutions, police or military forces. This is the key difference between rebellion and inter-communal violence which is conducted against ordinaiy citizens of some other ethnic groups within the same state. While the state authorities can be a neutral mediator between two ethnic groups in the inter-communal violence, the state authorities often become a party to the conflict in the case of rebellion.

This definition also implies the presence of some political purposes for the rebellion, whether they are related to self-determination, such as a greater autonomy or an outright secession/independence, or to the improvement of the

5 See, for example, Glazer & Moynihan 1975; De Vos & Romanucci-Ross 1982; Barth 1969; Despres 1975; Stack 1986. This question leads us also to vast literature in the study o f nationalism and its origins, which is also divided by similar approaches. See, for example, Smith 1986.

21 political, economic or social status of the ethnic group, or rather simply to the gain of political power in the country, such as the overthrow of the incumbent government.

Before the chapter proceeds to the review of the existing literature, a few additional notes are made here. The first one is on the usage of the word rebellion. This word sometimes entails negative connotations, which imply that violent actions that are being taken are illegal, illegitimate, unjustifiable, and so on. In analysing the separatism of Malay Muslims in Thailand, for example, one scholar noticed that the Thai state tends to refer the actions taken by the

Malay Muslims as “rebellion,” which implies a conflict involving a subordinate group within the state, while Malay-Muslims tend to call the same event the “war,” implying that what were in conflict were two distinct states

(Pojar 2005: 45). When I referred to the Serb actions in Bosnia as “rebellion (pobuna)” in an interview, a Croat historian Dr. Baric expressed his reservation in doing so because the Serbs were declared as “constituent nation” in Bosnia and in that sense they were different from the Serbs in Croatia.6 It must be underlined that the word “rebellion” in this thesis does not imply any inference about the legality, legitimacy, or justifiability of the actions taken by members of ethnic groups in any country. It is used as an objective and descriptive term.

Secondly, the analytic distinction between different types of violence does not mean that they are not related to each other. The occurrence of inter-communal violence or pogroms may precede the occurrence of rebellion: in Sri Lanka, for example, anti-Tamil violence by Sinhalese people occurred

6 Interview with Nikica Baric, Zagreb, 2005/11/08. Baric also told that there were some criticisms against the title of his book (Baric 2005) when he used the word “rebellion (pobuna)” for the actions o f Serbs in Croatia, because it has become a norm in Croatia to use the term “Great-Serbian Aggression (Velikosrpska A g resija )In this case, the word “rebellion” is disliked by the Croats because it implies the internal/domestic nature o f the conflict, whereas they tend to view the actions taken by the Serbs in 1991 as an “external” aggression.

22 before (but also after) the start of anti-state rebellion by the Liberation Tigers of Tamil Eelam (LTTE).7 In this case, the occurrence of inter-communal violence may be one of the causes of the occurrence of rebellion, leading to the perception by some members of the Tamil people that the current state would not protect them and they must establish their own state. Inter-communal violence may also occur after the anti-state rebellion. In Kosovo, for example, anti-Serb violence by the local Albanians has occurred even after the anti-state rebellion ended in 1999.8 The occurrence of rebellion often leaves behind an increased level of hostility between ethnic groups, and this may lead to the occurrence of inter-communal violence even after the end of rebellion. This so-to-speak “spill-over” effect of violence may be seen not only over time but also across space: an ethnic rebellion in one country, for example, may lead to the occurrence of inter-communal violence in another country. Dynamism between different types of violence, both within one country (over time) and across countries, is an interesting subject. It is however beyond the scope of the present thesis.

2. Four Problems: Critical Review of the Present Theoretical Literature

What are the causes of ethnic rebellion? Why do ethnic groups take up arms against the state? Various scholars have attempted to answer these questions in the past. However, there are some problems in the existing literature. In this section, four problems in the existing literature will be identified, namely (1) unitary actor assumption, (2) an unsolved controversy on

7 For the case of Sri Lanka, see for example Wilson 1988; Bose 1994. 8 For example, a large-scale anti-Serb and anti-UN riot occurred on 17 March 2004, which led to 19 dead, nearly 900 injured (out o f which 700 are Serbs), as well as destruction or damage of public buildings and Serbian churches. See, for example, ICG 2004.

23 the role of grievances, (3) lack of dynamics, and (4) lack of attention to the escalation of violence.

(1) Unitary Actor Assumption

The first problem in the present literature on ethnic rebellion is the unitary actor assumption. In many theoretical arguments and models that attempt to explain the occurrence of ethnic rebellion (whether they use the term

“ethnic rebellion” or not), it is assumed that ethnic groups act as unitary actors and make collective decisions to take up arms. One can see this assumption most vividly in a number of game-theoretic accounts of ethnic rebellion that have been developed since the beginning of the 1990s. According to the game theoretic accounts, the violence is explained by some strategic dilemmas or calculations made by the ethnic groups. While scholars emphasise different mechanisms to explain the violence, they all share the fundamental assumption that ethnic groups act as unitary actors, make some kind of strategic calculations and decide to take up arms at some point as a result of such calculations. In order to illustrate this point, let us briefly review some theoretical explanations in this field. These include (1) information failures, (2) commitment problems, (3) security dilemmas, and (4) signalling games.

According to Lake and Rothchild (1998: 11), information failure occurs when individuals and groups possess private information and incentives to misrepresent that information because revealing true information undercuts the ability of the group to attain its interest. When information failures occur, groups cannot acquire or share the information necessary to bridge the bargaining gap between them, and ethnic groups will decide to take up arms even though negotiation is actually possible. Information failures become more acute when the state weakens, since it is often the states that can communicate

24 and negotiate successfully between groups and thus help resolve information failures.

The concept of a “commitment problem” is proposed by scholars such as Fearon and Weingast. According to their explanation, ethnic conflict occurs because groups cannot make a credible commitment to uphold mutually beneficial agreements. Fearon (1998) argues that commitment problems arise from the differences in political power between groups. When one group is strong or growing stronger, this group cannot make a credible commitment to uphold mutually beneficial agreements since there is nothing to prevent it from breaking its promise in the future. As a consequence, a group that is weak or growing weaker has no incentive to believe their promises and would choose violent conflict over compromise. Weingast (1998) argues that the commitment problems arise from the uncertainty over the nature and intentions of other groups. According to him, one group may decide to take up arms if that group believes that there is even a very small chance that it may become a target of a genocidal attack.

Posen (1993) uses the concept of “security dilemma” for the analysis of violent ethnic conflict. According to him, the dilemma follows from the inability of the parties to know each other’s intentions directly. Therefore, even when one party expands its military capabilities strictly for defensive purposes, this can be taken as a threat to the other parties who in turn maintain and expand their military capacities. Because parties cannot know the intentions of others with certainty, what one does to enhance one’s own security causes reactions that, in the end, will make one less secure (Posen 1993: 104). When a spiralling arms race and hostility occurs as a result of security dilemma, it can lead to the pre-emptive attack: incentives to pre-empt arise when offence dominates defensive moves and thus the side that attacks first benefits

25 significantly from such a move. If there are strong incentives to use force pre-emptively, and if there is an uncertainty about the intentions of others, an ethnic group will make a decision to start military activities in order to avoid possibly worse results.

Finally, Oberg (2002) uses what he calls a “signalling game” to explain the occurrence of ethnic war. In his model, the occurrence of ethnic war is explained by a sequence of strategic decisions made by the ethnic group and the government: firstly, the ethnic group at some point makes a demand coupled with a threat of force that challenges government authority; the government then decides whether or not to stand firm and take countermeasures against the ethnic group; the ethnic group then decides whether it should act on its threats and stand up to the government or would back down; the government then decides whether to impose its authority with force or concede the issue at stake. The outcome of ethnic war occurs only when both the ethnic group and the government decide to use force rather than give concessions, and thus he argues that the theory must explain not only why the group rebels but also why the government resists by force. In his model, such factors as the expected value for war, value for the issues at stake, costs of backing down (making concessions) explain the decisions made by the ethnic group and the government.

As the brief overview above shows, the unitary actor assumption is quite common among game theoretic explanations of ethnic conflict. While these scholars use different models to explain violent ethnic conflict, they all assume (often implicitly) that ethnic groups act as unitary actors, make some strategic calculations and decide to take up arms at some point. However, such an assumption is problematic, because it is hardly a realistic assumption. There are often disagreements between actors within the ethnic group on the actions

26 they should take (or, on whether they should take some actions or not at all).

This may lead to the intra-ethnic competitions among actors, which has been analysed by some scholars.9 What is more important than the existence of intra-group disagreements itself is the fact that those who advocate a violent action, often called “radicals” within the ethnic group, can freely do so without any formal decision-making or any intra-ethnic consensus within the ethnic group. These radicals may decide to take up arms even though many people in the ethnic group think that it does not serve the interest of the group.

Indeed, some authors have pointed out that the unitary actor assumption is problematic and that it constitutes a weakness of the theoretical arguments mentioned above (e.g. Brubaker & Laitin 1998). What has not been argued and explored so far, however, are the consequences of relaxing this assumption. After all, assumptions made in theoretical models often contain some kind of unrealistic element: theory always entails simplification of the reality and thus naturally less complex than the reality itself. If relaxing the assumption does not change our view on causes and dynamics of the conflict very much, there is no point in relaxing the assumption. What are the consequences of relaxing this assumption? What are the implications for the study of causes of violent ethnic conflict and for the policy recommendations to prevent the occurrence of violent conflict? These questions largely remain unanswered so far. This thesis thus attempts to address these questions through comparative case study and large-N analysis.

(2) A Controversy - Do Grievances Matter?

The second problem is related to a controversy in the present literature

9 For example, see a series of analyses on what is called “ethnic outbidding,” e.g. Rabushka & Shepsle 1972; Hislope 1996.

27 on the role of grievances for the occurrence of ethnic rebellion. On the one hand, many scholars have argued that the grievances held by the members of ethnic groups are important since the grievances motivate them to take up arms, and thus are one of the causes of ethnic rebellion. On the other hand, some scholars have presented a controversial argument that “grievances do not matter” based on the results of their large-N analyses.

Traditionally, many scholars have argued that the grievances are important. This is not limited to the field of political science. In the field of social psychology, for example, Tedeschi and Nesler (1993) use “grievances” as an intervening variable that explains the violent actions of individuals.

According to them, the aggrieved individual resorts to aggression or violent attacks when he or she fails to settle a “grievance” formed because of some negative events which an outside party is blamed for.

Among political scientists, Gurr argues that grievances do matter for the occurrence of ethnic rebellion, even though other factors are also important. In his recent study, Gurr (2000: 65-95) presented a general framework for explaining and analysing ethnopolitical rebellion, which consists of four general categories of factors, namely (1) salience of group identity, (2) group incentives for collective action, (3) group capacity for collective action, and (4) domestic and international opportunities. As for the role of grievances, Gurr

(2000: 163-164) argued that “collective disadvantages are the root cause of ethnopolitical action” and “redress of grievances about invidious treatment and the desire to gain advantages are principal sources of group incentives for action.” He conducted a large-N analysis using the Minorities at Risk (MAR) dataset, which he constructed with his colleagues, to support his argument.

Dudley and Miller (1998), using the MAR dataset, also found that two types of grievances, namely political discrimination and lost autonomy, have a direct

28 significant effect on the occurrence of ethnic rebellion.

Contrary to these arguments, however, some scholars have reached to the conclusion that “grievances do not matter.” This argument has been originally put forward by an economist Collier and his colleague Hoeffler

(Collier 1999; Collier & Hoeffler 1999; Collier & Hoeffler 2001). Based on the results of large-N analysis, he argues that the objective indicators of

“grievance” do not explain the occurrence of civil war and that indicators related to “greed” and “opportunity,” such as primary commodity export, male secondary education enrolment, per capita income and growth rate, have a significant effect on the onset of civil war. Fearon and Laitin (2003) also supported this conclusion, arguing that there is no evidence for “grievances” as a cause of civil war. According to Fearon and Laitin, what matters is the presence of the “right conditions of insurgency,” such as a weak state, mountainous terrain and instability in the centre. While their analyses are on the occurrence of civil war in general and not confined to ethnic conflict, the cases of civil war analysed by these authors include ethnic civil wars, and thus their explanation applies to ethnic rebellion as well.

Which camp is right in this controversy? At least as far as ethnic rebellion is concerned, one can argue that the controversy is not fully resolved, since both camps have some problems in their research designs and thus one cannot draw a definitive conclusion from the existing literature.

Firstly, the results of studies of Gurr and other scholars using the

MAR data are not totally convincing, since their research design fails to include the variables related to the opportunities identified by scholars who argue that “the grievances do not matter.” For example, Gurr does not even include a variable which shows the most powerful and consistent results, namely the level of economic development (GDP per capita). He uses only the

29 indicators of political system (democratic polity) and regime instability as variables related to domestic opportunity (Gurr 2000). The variables that relate to the opportunity structure for ethnic rebellion have not been fully included in the analyses using the MAR dataset, and they thus face a potential problem of omitted variable bias.10

Secondly, the results of analyses of those who argue that “grievances do not matter” are not fully convincing for the students of ethnic conflict for several reasons. Firstly, there is an implicit assumption that civil wars are homogenous: by treating all types of civil war equally as events that are explained by a single set of explanatory variables, these scholars implicitly assume that the events coded as an occurrence of “civil war” are homogenous and have the same causes. Some scholars criticise this assumption. For instance,

Sambanis (2001: 265) argues that “the new economic theories of civil war do not consider if different war types have different causes, and their research designs, which aggregate all civil wars in a single category, implicitly suggest that there are no such differences.” The conclusions drawn at the aggregate level, therefore, may not be valid when one limits his analysis to the cases of ethnic rebellion. The second reason is the unit of analysis. In the large-N analyses of the occurrence of civil war, the unit of analysis is a country-year, and variables specific to the ethnic groups cannot be included in the statistical analyses. The factors that are found significant in the analyses using the MAR dataset, therefore, were not included in the analyses conducted by the authors who denied the importance of the grievances. While the conclusions drawn by the two camps are different, this difference thus might be an artefact of the different ways of operationalising “grievances.”

A controversy between the two camps about the role of “grievances”

10 For the problem o f omitted variable bias, see King et al 1994:168-182.

30 is not solved, and this thesis attempts to find an answer to this controversy. In order to do so, it is necessary to conduct a large-N analysis using both the

MAR dataset and the opportunity-related variables that were found to be statistically significant in the earlier works on the onset of civil war.

(3) Lack of Dynamics

The third problem in the present literature is a lack of dynamics. Many of the theoretical arguments and models remain static. As Oberg (2002: 14-19) pointed out, most explanations for ethnic conflict are structural and focus primarily on the objective conditions that affect the decision by ethnic groups to take up arms against the government. When it comes to this point, scholars mentioned in the previous section, such as Gurr, Collier, Fearon and Laitin, share the same premise despite the different conclusions they reach on the role of grievances: both camps assume that the structural conditions affect the decision by the rebels to take up arms, and there is no dynamic element in their theoretical models.

This lack of dynamics is particularly visible in the theoretical arguments based on the large-N analyses, even though the tendency to focus on the structural conditions can be observed among scholars who do not conduct large-N analyses as well.11 It is because the research design of large-N analysis is based on the very assumption of structural determination: conditions or factors that affect the occurrence of ethnic conflict in the theoretical arguments are assumed to be independent from and exogenous tothe dependent variable.

Indeed, one will not be able to conduct a standard large-N analysis (such as regression analysis) once one assumes that the explanatory factors are not

11 Brown, for example, pointed out that the literature on internal conflict had primarily focused on the underlying factors or permissive conditions. See Brown 1997: 4.

31 independent and exogenous.

While this thesis does not deny the validity of the structural determination, it will attempt to make a model more dynamic in two ways as a remedy to the problem of lack of dynamics. One way to make a model more dynamic is to consider an interaction between the rebels and the state authorities against which the rebels take up arms. For example, O’Leray and

Tirman (2007: 12) criticised that the econometric analyses of civil war are

“analytically one-sided,” asking “why are states - rather than just insurgents - not modelled as suppliers of violent exploitation or predation, a perspective that would seem more consistent with the genre of economic theorizing from which their work flows?” Oberg’s recipe for the problem of lack of dynamism is to “bring the government’s calculations back into the equation, and make the argument strategic” (Oberg 2002: 18). Indeed, more than 40 years ago,

Eckstein (1965: 145) pointed out that it is important to examine the role of the state authorities in explaining internal war: “One crucial choice that needs to be made is whether to put emphasis upon characteristics of the insurgents or incumbents, upon the side that rebels or the side that is rebelled against. Not surprisingly, the existing literature concentrates very largely on the rebels...

This would seem to be only natural: after all, it is the rebels who rebel. At least some writings suggest, however, that characteristics of the incumbents ... must be considered jointly with characteristics of the insurgents, indeed perhaps even emphasized more strongly.” His argument still remains valid and applicable to the study of ethnic rebellion. While he presented it as a “crucial choice,” it is not an either-or choice: one can, or indeed should, consider both sides to fully explain the occurrence of ethnic rebellion. This thesis thus attempts to bring the state reactions back in the theoretical arguments, while also paying attention to the conditions that motivate and enable the rebels to

32 take up arms.

Another way to make a model more dynamic is to take time more seriously. Theoretical models based on the idea of structural determination are fundamentally static, because it is not assumed that the time (duration of some events, for example) affects the occurrence of ethnic rebellion. However, this becomes important when one conceptualises “grievances.” In many theoretical arguments based on the large-N analyses, the grievances are operationalised by some objective indicators (such as gini coefficients that measure the degree of economic inequality in the society), and thus it is assumed that the level of grievances will remain constant over time as long as the structural conditions do not change. However, such static conceptualisation of grievances may not fully capture the dynamic element of grievances: people may become more and more frustrated as the time goes by, even if the situation does not change (and thus the objective conditions do not change). Indeed, people may be more frustrated precisely because the things do not change (and thus things do not get better). The following chapters, therefore, will attempt to conceptualise a dynamic element of grievances.

(4) Lack of Attention to the Escalation of Violence The fourth problem in the present literature is the lack of attention to the escalation of violence. This is visible both in the game-theoretic accounts of ethnic rebellion and the large-N analyses of civil war. In the large-N analyses of the occurrence of civil war, for example, the dependent variable is binary (0 or 1), and thus the occurrence of violence is conceptualised as dichotomous: there is a peace (0), or civil war (1). Under such a research design, the occurrence of civil war is inevitably conceptualised as a one-shot event. The peace lasts until some year (y=0 while t=0, 1, .... n), and then the

33 civil war occurs in the next period (y=l when t=n+l). In a statistical analysis where the “occurrence” of civil war is regressed on the explanatory variables, therefore, the objective conditions observed in the year of the “occurrence” will be associated with the onset of civil war and will thus be judged as “causes” of civil war. This conceptualisation of large-scale violence as a one-shot event is also seen in the game-theoretic accounts of ethnic rebellion. The unitary actor assumption on ethnic groups discussed above affects this conceptualisation.

Recall that the ethnic group is assumed to be a unitary actor and it is assumed to take a collective decision to take up arms at some point. The occurrence of large-scale violence is, therefore, a one-shot event, which occurs when the ethnic group makes a decision. The peace lasts until the very moment when the ethnic group decides to take up arms, and the war occurs once such a decision is made by the ethnic group.

This conceptualisation may be less problematic in the case of international warfare, because the state normally has a formal decision-making body, and a large amount of financial, human and military resources will be instantly mobilised once the decision is made by the government to go to war.

One can thus draw a line between peace and war, and can assume that the occurrence of war is a one-shot event, caused by the decision of the government to go into war. This conceptualisation becomes more problematic, however, when it comes to ethnic rebellion or indeed to internal war in general, because the violence can be initiated by a handful of radicals, and can start from very low intensity, such as sporadic bombing attacks. If the violence starts from a low level, there must be an escalation of violence when a large-scale conflict occurs. The theory that attempts to explain the occurrence of large-scale violence, therefore, must answer two distinct questions, namely (1) why do these radicals take up arms in the first place? and (2) why does a

34 low-intensity conflict escalate into a large-scale one? It may well be the case that the factors that explain the occurrence of low-intensity rebellion would not explain the escalation of violence.

The possibility that the occurrence of large-scale violent conflict can be preceded by the occurrence of low-intensity violence is particularly problematic for the large-N analyses of civil war, because this possibility suggests a potential problem of misidentification of structural causes of the conflict. If the occurrence of large-scale conflict is a joint outcome of the occurrence of the low-intensity conflict and its escalation into large-scale one, the factors that cause the low-level conflict should be considered as part of the causes of conflict, but these factors are not captured by the statistical analysis that only takes into account the moment when the conflict become a large-scale one. In addition, the occurrence of the low-intensity violence may even cause the occurrence of escalation: indeed, Part IV of this thesis will argue that the occurrence of low-intensity rebellion tends to incite military repressions by the state authorities and the repression in turn tends to cause the escalation of rebellion. Then, one cannot simply ignore the occurrence of low-intensity violence even when one attempts to explain the occurrence of large-scale conflict, while scholars conducting large-N analyses of civil war would simply assign the value of 0 as long as the violence remains a low-intensity one.

This does not mean, however, that all large-scale conflicts start from low-intensity violence. In some cases, a large-scale conflict may well occur suddenly without much escalation of violence involved. Indeed, the present thesis will show that some ethnic rebellions start from low-intensity violence while others are large-scale ones from the outset, and this variance itself is interesting and requires a theoretical explanation.

Because of the lack of attention to the escalation of violence, the

35 mechanisms that account for the relationship between the low-level violence and the high-level one have been mostly ignored in the existing literature.

Examining such mechanisms, therefore, will be a useful contribution to the understanding of the occurrence of large-scale ethnic rebellion.

3. Three-Step Approach for the Analysis of Ethnic Rebellion

The critical review of the existing theoretical literature conducted above suggests the necessity to take a three-step approach for the analysis of ethnic rebellion. In this section, this three-step approach is discussed.

(1) Initial Level of Rebellion

Firstly, one must identify the initial level of rebellion, because this suggests the different tasks for explaining the occurrence of large-scale ethnic rebellion. If the ethnic rebellion starts from low-intensity violence, one should firstly explain why the low-intensity violence occurs in the first place, and then explain why the low-intensity violence escalates, or does not escalate, into a large-scale one. If the initial level of rebellion is low, therefore, one should explain (1) the onset of low-intensity rebellion, and (2) the escalation of the rebellion into a large-scale one. If the ethnic rebellion immediately starts on a large-scale, on the other hand, one should explain why the rebellion occurs, and this should suffice as an explanation of the occurrence of large-scale ethnic rebellion. The relationship between the initial level, onset of the initial rebellion, escalation and the final outcome (“onset” of large-scale rebellion) is presented in Table 1.1.

36 Table 1.1 Initial Level, Onset and Escalation

Onset of Initial "Onset" of large-scale Initial Level Escalation Rebellion Rebellion

Low Yes Yes Yes (1)

Low Yes No No (0)

Low No - No (0)

High Yes - Yes (1)

High No - No (0)

This table shows how the classic binary coding of “onset” of large-scale conflict mixes outcomes of different dynamics into the same category. On the one hand, the onset of large-scale conflict (normally coded as

1) can be either an outcome of the onset of high-intensity rebellion at the first place or a joint outcome of the onset of low-intensity rebellion and the occurrence of escalation. On the other hand, the non-occurrence of the large-scale conflict (coded as 0) can mean either lack of the initial rebellion at all (whether it is high or low intensity) or the occurrence of low-intensity rebellion which does not experience an escalation. By taking into account the initial level of rebellion, therefore, one can conduct a more nuanced analysis of the occurrence of ethnic rebellion.

The attention to the initial level of rebellion naturally raises one question: why do some ethnic groups experience an onset of low-intensity rebellion while others experience an immediate onset of large-scale rebellion?

Because of the lack of attention to this aspect, this question has been mostly ignored in the existing literature. In order to answer this question, one must examine who actually decide to take up arms among the ethnic group, and what

37 are the resources available to them. This will be the task of Part II.

(2) Onset of the Initial Rebellion

The second step for the analysis of ethnic rebellion is to analyse and explain the onset of the initial rebellion. Whether the initial level of rebellion is low or high, the large-scale conflict cannot occur without the occurrence of initial rebellion. Explaining the occurrence of initial rebellion is, therefore, a central task for the explanation of ethnic rebellion. To fulfil this task, one should analyse why those who took up arms actually decided to do so. This will be the task of Part III. The controversy on the role of “grievances” in the existing literature is related to this question, and an answer to this controversy will be sought in Part HI as well. The attention to the initial level of rebellion discussed above raises one question related to the second step: do the low-intensity rebellion and large-scale rebellion have the same causes or not? Because the issue of initial level of rebellion itself has not been addressed much in the existing literature, most analyses on the causes of the rebellion ignore the difference between the onset of low-intensity rebellion and high-intensity rebellion. Therefore, Part HI will attempt to examine whether the onsets of low-intensity rebellions and large-scale ones have the different causes or not.

(3) Escalation of the Rebellion

The third step for the analysis of ethnic rebellion is to examine the dynamics of escalation of the rebellion from the low-intensity one to the large-scale one. The onset of the low-intensity rebellion does not necessarily mean that it will escalate into large-scale one: some will do so while others remain low-intensity ones or may de-escalate without experiencing an

38 escalation into large-scale conflict. What are the causes and determinants of the occurrence of escalation? Part IV will attempt to examine the mechanisms of escalation and to answer this question. This task is, of course, limited to the cases where the rebellion starts from a low-intensity one, because by definition one will see the escalation of rebellion only when it starts from a low-intensity one.

Examining the mechanisms of escalation is particularly important for learning more nuanced lessons for the prevention of large-scale conflict. The analysis based on the binary coding of the occurrence of large-scale conflict will only suggest how to prevent the occurrence of conflict at all. The three-step approach adopted in this thesis will suggest that there are actually two steps to prevent the occurrence of large-scale conflict. Firstly, one can attempt to prevent the occurrence of the initial rebellion. In addition, however, the three-step approach suggests that one can attempt to prevent the escalation of rebellion if the initial level of rebellion is low. This latter point cannot be learned as long as one adopts the binary conception of the occurrence of large-scale rebellion, and this is where the present thesis can possibly make some contributions to the study of conflict prevention.

Conclusion

This chapter has attempted to present the theoretical premises for the thesis. This chapter firstly defined the key term for the thesis, “ethnic rebellion.” It then presented a critical review of the theoretical works in the existing literature, identifying four major problems. It finally presented a three-step approach for the analysis of ethnic conflict, which would examine

(1) initial level of rebellion, (2) onset of the initial rebellion, and (3) escalation.

39 The present thesis will take this three-step approach in analysing the causes and mechanisms of ethnic rebellion. Before it embarks on the empirical analysis, however, the next chapter will address the methodological premises for the present thesis.

40 Chapter 2 Methodological Premises: Large-N analysis and the ex-Yugoslav Case Study

Introduction

The present thesis adopts a “mixed methods” approach that combines large-N analyses and in-depth case studies. After explaining why it is needed, this chapter briefly discusses the MAR dataset which will be the basis for the large-N analysis in this thesis. It will then discuss the four cases in the ex-Yugoslav region for in-depth case studies. Finally, it will present the structure of the rest of the thesis, which reflects the three-step approach discussed in Chapter 1 and the mixed methods approach discussed in this chapter.

1. Methodology: Why Mixed Methods?

The present thesis adopts a mixed methods approach1 and attempts to conduct both relatively intensive case-studies and large-N analyses to support the theoretical arguments. A reader may ask, “why mixed methods?” Indeed, the usage of the mixed methods has been relatively rare in the study of ethnic conflict. On the one hand, a vast volume of studies on ethnic conflict have been produced without any statistical analysis. Some studies even mention an extremely large number of cases qualitatively without resorting to statistical techniques.2 On the other hand, when the large-N studies are conducted, these

1 For the excellent review of the discussion on the mixed methods in political science, see Lieberman 2005. See, for example, Horowitz 2001, which is based on the reports of “approximately 150

41 studies often lack a qualitative analysis at all. Oberg (2002: 77), for example, decided not to conduct any intensive case study because case study methods have difficulties in handling probabilistic propositions. Even when the qualitative case study is combined with the large-N analyses, the former is often treated as a mere illustration, and thus is not considered as an integral part of the original research. Gurr (2000), for example, includes a few pages of descriptive “sketches” on a total of 14 communal groups in the world, but these sketches are written by different authors (experts on each case) in the MAR project and are far too brief to be considered as an original analysis of these cases.

In the present thesis, mixed methods are adopted for several reasons. As for the large-N analysis, the theoretical aspiration of the thesis requires it.

As discussed in Chapter 1, this thesis has identified some problems in the existing literature, and it attempts to develop theoretical arguments that address these problems. In order to generalise in a convincing manner, it is necessary to conduct some kind of large-N analysis, due to the potential problem of selection bias. As Geddes has rightly pointed out, the question of how far one can generalise the argument cannot be answered as long as one sticks to a few cases; one can only answer that question when one broadens the scope of the analysis sufficiently.3 Another problem with the small-N case study is identified by Lijphart (1971: 685) as the problem of “many variables, small number of cases.” A strength of the large-N analysis relative to the small-N analysis is “its ability to simultaneously estimate the effects of rival riots in 50 countries” (p. 29). For criticism of Horowitz on methodological grounds, see Laitin 2001b. 3 Geddes 1990; Geddes 2003: 89-129. Of course, this doesnot mean that one cannot make a generalised theoretical argument without conducting large-N analyses. One can create or propose a general theoretical argument based on single-N or small-N studies and one can even test the hypotheses with a qualitative case study. See, e.g., Van Evera 1997: 55-67; Lieberman 2005: 442-443.

42 explanations and/or control variables on an outcome of interest” (Lieberman

2005: 438).

Case studies are useful, however, because large-N methods also have some weaknesses. As is often pointed out, correlation and causation are not the same. Goldthorpe (2001: 2) pointed out, for example, that it is “widely recognized in both philosophy and statistics” that “correlation - or, more generally, association - does not imply causation” even though “causation must in some way or other imply association.” The statistical methods do confirm the existence of correlation, but not necessarily causation. It is the theory which clarifies the causal mechanisms that (are assumed to) underlie the association.

Let us quote Goldthorpe (2001: 14) again: “Causal explanations cannot be arrived at through statistical methodology alone: a subject-matter input is also required in the form of background knowledge and, crucially, theory.” The crucial task is, therefore, to make a theory convincing. Here, the case study can play a significant role. One of the strengths of the case study is that one can conduct “process-tracing,” i.e. exploring “the chain of events or the decision-making process by which initial case conditions are translated into case outcomes” (Van Evera 1997: 54-55, 64-67), and thus can unwrap the cause-effect link in the real cases. Intensive case-studies thus can make generalised arguments empirically more convincing. As Van Evera argued,

“large-/? methods tell us more about whether hypotheses hold than why they hold. Case studies say more about why they hold.”4

4 Van Evera 1997: 55 (emphasis added). Of course, not all scholars would conduct a case-study to make a theory more convincing. Others would say that they can use formal modelling to make a theory logically convincing and internally consistent. See, for example, Laitin’s “tripartite methodology” which consists o f statistics, formalization and narrative (Laitin 2002). The strength of the formal modelling is its ability to provide “an internally consistent logic that accounts for the stipulated relationships among abstract variables” and assure us that “our causal stories are coherent and noncontradictory.” See Laitin 2002: 631. One may still argue, however, that the assurance of the internal consistency does not guarantee that it is empirically convincing.

43 This thesis also aspires to make some contribution to the accumulation of knowledge in the area study of the ex-Yugoslav region by conducting a theory-driven “controlled” comparative analysis. As some scholars have pointed out, “controlled comparisons” have been relatively few in the study of ethnic conflict in general. For example, Varshney (2001, 2002) argues that there has been too much attention on the cases of serious ethnic conflict and that analysts have not explored the factors, conditions or variables that explain both the occurrence and non-occurrence of the conflict. Brubaker and Laitin

(1998: 435) also pointed out that “controlled comparisons have been relatively few, especially those comparing regions suffering from ethnic violence with regions in which similar ethnic conflicts have not issued in violence.” As for the ex-Yugoslav region, Brubaker and Laitin (1998: 436) suggested that the

“breakup of Yugoslavia has most often been treated as a single complex interconnected case, but if we had adequately disaggregated data, it could be studied as a set of cases.” The present thesis is indeed an attempt to move toward this direction, i.e. to treat cases of ethnic rebellion in the ex-Yugoslav region separately and to explore commonalities and differences between the cases. By doing so, this thesis attempts to contribute to the further and clearer understanding of the dynamics of ethnic conflict in the ex-Yugoslav region.

2. The MAR Dataset

This section briefly discusses the MAR dataset, which will be used for the large-N analyses in this thesis. According to the codebook of the MAR dataset, the MAR Project is “an independent, university-based research project that monitors and analyses the status and conflicts of politically-active communal groups in all countries in the world with a current population of at

44 least 500,000” and it is “designed to provide information in a standardized

format that will aid comparative research and contribute to the understanding

and peaceful accommodation of conflicts involving communal groups.”5 The

MAR dataset contains various kinds of information on over 270 ethnic groups

in the world that are judged as being “minorities at risk.” According to the

codebook of the MAR dataset, a “minority at risk” refers to “an ethnopolitical

group (non-state communal group) that (1) collectively suffers, or benefits

from, systematic discriminatory treatment vis-a-vis other groups in a society;

(2) and/or - collectively mobilizes in defence or promotion of its self-defined

interests.”6 Since the MAR project was initiated by Ted Robert Gurr in 1986,

some ethnic groups have been added and others have been eliminated from the

dataset, based on the judgments made by the group of scholars of the MAR project.

One should pay attention to the presence of this selection process,

because there are some questions that one cannot address by using the MAR dataset due to the selection process. For example, it is assumed that the

communal groups included in the MAR dataset have already developed their

communal or ethnic identity. This assumption in turn prohibits us from

addressing some questions such as the following: why and how do some

groups develop their own ethnic identity while others fail to do so? Note that

this is one of the most important questions asked in the study of nationalism

and ethnicity. Since it is methodologically assumed that the communal or

5 The codebook of the MAR dataset, as well as the qualitative and quantitative data constructed by the MAR project team, can be downloaded from the website of the MAR project at the Center for International Development and Conflict Management (CIDCM) and the University of Maryland at: http://www.cidcm.umd.edu/inscr/mar/index.htm. 6 See the codebook of the MAR, “Minorities at Risk: Dataset Users Manual, 030703” available at http://www.cidcm.umd.edu/inscr/mar/index.htm. See also Gurr 2000: 7-13, on the definition o f ethnopolitical groups “at risk” and the operational rules for the inclusion o f the groups in the dataset.

45 ethnic identities already exist, one simply cannot address the question of the

origin of communal and ethnic identity by using the MAR dataset.

One should also pay attention to the way the variables are coded. The

overwhelming majority of the variables included in the MAR dataset are coded

as binary or ordinal categorical variables. Furthermore, the coding of the

variables is far from easy and automatic, requiring careful qualitative judgment

by the specialists on the area. Let us take one example: a variable ECPOV,

“Severity of economic disadvantages (1990-95)” is coded according to the

following scale: 0 - No disadvantages; 1 - Slightly disadvantaged; 2 -

Substantially disadvantaged; 3 - Very seriously disadvantaged; 99 - No basis

for judgment.7 What is the difference between “substantially” disadvantaged

and “very seriously” disadvantaged? The criteria are not very clear from the codebook, and the same fact (presence of a certain level of disadvantage) can

be coded as 1, 2, or 3 according to the qualitative judgment of the person(s)

who code specific cases. Even when the absolute number is the basis of judgment, the coding can be equally difficult, because it is sometimes difficult

to establish a “fact” itself in a conflict situation. Let us take an example of a

REBEL variable (Rebellion Index). According to the codebook, the rebels must

have more than 1,000 armed fighters to be coded “6” of the rebellion score. But

it is sometimes difficult to obtain correct information, since rebels often tend to

exaggerate the strength of their forces. In some cases, the reliable data on the

number of fighters may not be found at all.

Indeed, the MAR project group admitted that they have “not yet

assessed inter-coder reliability nor have its indicators been systematically

screened for internal consistency.”8 They claim that coders are “well-trained”

7 See the codebook of the MAR: 39. 8 See the codebook o f the MAR: 8.

46 students who “undergo a rigorous training period,” that all coding is reviewed

by senior personnel and that “overall the assessment of coding reliability and

validity is satisfactory.” 9 They however admitted that “tests of internal

consistency by Matthew Kocher of the University of Chicago in spring 1999

identified some problems with coding,” even though “the MAR staff

immediately moved to address these concerns.”10 Since the coding of the data

relies heavily on the qualitative judgment, which would be extremely difficult

at least in some cases, one should note that the problem of measurement error

could be potentially serious when one uses the MAR dataset.

To address the potential problems of the MAR dataset, some authors

have made alterations when they use the MAR dataset for their analysis. Oberg

(2002: 94), for example, added a large number of ethnic groups to the MAR

dataset to address “potential problems with sample selection bias built into the

Minorities at Risk data set” and to increase a number of ethnic groups from 285

to 653. Metz (2005: 6) used the MAR dataset to select cases when he made a

list of rebellion episodes, but then he conducted an extensive “content

analysis,” using six secondary sources, to (a) confirm that the cases in the

dataset represent actual rebellions, and (b) to confirm that the rebellions were

indeed ethnic. After a content analysis, he indeed found that 24 cases that were

originally in the MAR dataset were discarded either because of the absence of

evidence that any rebellion had taken place or because they were anti-colonial

rebellions (Metz 2005: 13). The present thesis also conducted a survey on the

organisations that were behind the first onset of rebellion activities (see

Chapter 3 for the details) and indeed found that a few cases are difficult to regard as an episode of “rebellion” as defined in the previous chapter. Let us

9 See the codebook o f the MAR: 8. 10 See the codebook o f the MAR: 8.

47 take an example of the Vietnamese in Cambodia. For this ethnic group, a rebellion score (REBEL) takes a value of 1 in the year 2000. What happened in

2000 in Cambodia, as far as the author could discover, was that “a group of over 20 ethnic Vietnamese fisherman [sic] killed 3 government fisheries officials in Kandal province and wounded 3 others during a late-night government raid on illegal fishing practices on the Tonle Sap river” (BDHRL

2001). While it was indeed committed by members of the ethnic group against the government officials, this act does not seem to have been committed by an ethnic organisation with a political purpose, and seems rather a spontaneous reaction against the threat of arrest. If this is true and there was no other episode of violence in the same year, it cannot be regarded as “rebellion” at least according to the definition in the present thesis.

In the present thesis, however, an adjustment or re-coding of the MAR dataset is not conducted. As for the issue of sample selection, what is important is the method of sample selection rather than the number of cases. Therefore, simply adding more cases to the dataset does not solve the question of the selection bias caused by the sampling method (if any) unless the addition of the cases is conducted in such a way that it can remedy the problem. Otherwise, the addition of more cases can be even problematic, since this may indeed favour the argument of the person who adds the cases. Without identifying a problem in the sampling method and a remedy to solve the problem, therefore, one should not simply add cases to the original dataset. As far as the present thesis is concerned, the sampling method does not particularly favour the argument made in the thesis, and thus does not pose a serious challenge against the inferences made based on the MAR dataset.

A check on the validity of the coding was not conducted for a rather practical reason. If a validity check should be conducted, it should be

48 systematic, based on clear criteria, and comprehensive, covering all relevant observations, since a partial and chaotic correction of the dataset would simply compound the problem of the inter-coder reliability. Note that the present thesis deals with the onset as well as escalation of the rebellion of all ethnic groups included in the dataset in the period from 1945 to 2000. An attempt to check the validity of the coding of the key dependent variable, a rebellion index

(REBEL), would involve checking the validity of the coding of 1442 instances across space and time. It is practically impossible to conduct such a large-scale survey alone in a reasonable time. The MAR project can mobilise far more financial and human resources for the coding.

It was thus decided to use the MAR dataset as it is. The large-N analysis conducted in the present thesis, therefore, depends on the MAR dataset, even though it also uses some variables constructed by the author from scratch (Chapter 3) and some variables taken from other sources. If the theoretical arguments are supported by some large-N analysis, therefore, this suggests that the theoretical arguments are valid as far as the ethnic groups and countries covered by the MAR dataset are concerned, assuming that the perception and the qualitative judgments of the scholars (coders) in the MAR project are right and consistent across space and time.

3. The ex-Yugoslav Cases: Croatia, Bosnia, Serbia and Macedonia

The present thesis will analyse four cases of recent ethnic rebellions in the ex-Yugoslav region, namely the Serb rebellions in Croatia (1991-) and

Bosnia (1992-) and the Albanian rebellions in Kosovo (Serbia, 1996-) and

Macedonia (2001). In addition to the fact that the controlled comparisons that treat these cases separately have been relatively few as Brubaker and Laitin

49 observed (see p. 44 of this thesis), there is one strong reason for conducing comparative analyses of these four cases: while these four cases all occurred in the overall context of the dissolution of ex-Yugoslavia, these cases show remarkable differences when it comes to the subject of the present thesis.

Firstly, one can observe a difference in the initial level of rebellion: the initial level of rebellion was very high in the case of the Serb rebellions in both

Croatia and Bosnia, while it was quite low in the case of the Albanian rebellions in both Kosovo and Macedonia. Secondly, when one compares the two Albanian rebellions in Kosovo and Macedonia, one also finds a difference between these two cases: while the Albanian rebellion in Kosovo escalated into a large-scale one, the Albanian rebellion in Macedonia remained a relatively low-intensity one and then de-escalated after a peace agreement. For the present thesis, therefore, there is a variance in the dependent variables among these four cases, which make it possible and interesting to conduct in-depth comparative analyses. In this section, a brief overview of the four cases is presented as an introduction to the case studies.

In Croatia, the Serbs constituted a sizeable minority of the population

(12% in 1991), while the Croats constituted an absolute majority of the population (78% in 1991).11 Croatia had its first democratic elections in April-

May 1990, and the opposition party led by Franjo Tudjman won the absolute majority of the seats in the parliament and established a new government

(Tudjman was elected as the new President of the Republic by the newly-constituted parliament). 12 After its inauguration, the Tudjman government started implementing a series of nationalistic policies, such as the introduction of a new national flag, coat of arms and police uniform, the

11 For the census results in Croatia from 1961 to 1991, see Woodward 1995: 33. 12 For the details of the first democratic elections in Croatia, see Grde§i6 et al 1991.

50 constitutional amendment that made the usage of Latin script obligatory, and the adoption of a new constitution that defined Croatia as “the national state of the Croatian nation.”13 At the same time, the Tudjman government, together with the Slovenian government, pursued the policy of independence from the

Yugoslav Federation. Against these policies, the Serb politicians started to organise the Serbs in Croatia, to create a Serb territorial unit within Croatia and to claim its secession from Croatia. They started preparing for the rebellion against the Croatian government, and the full-scale rebellion was started after the declaration of sovereignty by the Croatian government on 25 June 1991.

In 1991, Bosnia was a multi-ethnic republic with three “constituent nations,” namely Muslims (43.7% in 1991), Serbs (31.4%) and Croats (17.3%).14 Bosnia had its first democratic elections in November-December

1990, and the three newly-established nationalist parties that represent three ethnic groups won the elections.15 These three parties formed a coalition government and established a power-sharing regime after the elections. This coalition government, however, collapsed in October 1991, when a

“memorandum on the sovereignty of Bosnia” was adopted in the parliament while the Serb politicians opposed it and left the parliament before the voting.16

After this event, the Serb politicians started its centrifugal actions, establishing its own institutions (such as the Serb Parliament and the Bosnian Serb

Republic) and conducting a Serb referendum on the decision to remain in the

Yugoslav Federation. At the same time, the Serb politicians started preparing

1 For the details o f the political developments after the elections in Croatia, see e.g. Silber & Little 1996; Cohen 1993; Glenny 1996. 14 For the census results in Bosnia from 1961 to 1991, see Woodward 1995: 33. 15 For the details of the first democratic elections in Bosnia, see Amautovic 1996. 16 For the details o f the adoption of “memorandum” and the political developments after that, see e.g. Burg & Shoup 1999: 76-127. A full text o f the “memorandum” as well as the counter proposals by the other parties in the Bosnian Parliament can be found inBorba, 1991/10/16.

51 for the rebellion, and the full-scale civil war started after the Bosnian parliament declared the independence of Bosnia on 3 March 1992 based on the results of the national referendum on the independence of Bosnia.

Under the constitutional arrangement of the communist regime, Serbia had two “autonomous provinces,” Vojvodina and Kosovo, within its territory.

While the Serbs constituted an absolute majority of the total population of the republic of Serbia (65.8% in 1991) and the Albanians constituted a significant minority (17.2%), the Albanians constituted an absolute majority within the autonomous province of Kosovo (about 90% in 1991) and the Serbs constituted a minority within Kosovo (about 10%). Large17 demonstrations by the Kosovo

Albanians occurred in 1968, after the downfall of a conservative leader Rankovid in 1966 who had firmly rejected their request for autonomy (Lampe

1996: 296). The Yugoslav federal authorities made a series of concessions to the Albanians after 1968, including the increase of the Albanian officials in the provincial authorities, right to raise the Albanian flag, permission to establish a university with instruction in the Albanian language, and so on (Pipa 1989).

Republics and Autonomous Provinces became almost fully equal under the

1974 Federal Constitution, according to which the two autonomous provinces gained full autonomy over their parliaments, budgets and judicial systems, and the governments of the provinces could veto any policy of the Republic of

Serbia while the Serbian government did not have equivalent powers over decisions made by the provincial governments.18 The Kosovo Albanians, however, again waged demonstrations in 1981 after the death of Tito, after which the radical Albanians were purged from the party or arrested by the authorities.19 The inter-ethnic relations in Kosovo started to deteriorate

17 For the census results in Serbia from 1961 to 1991, see Woodward 1995: 34. 18 Woodward 1995: 40,65. See also Vickers 1998: 178-181. 19 As for the 1981 demonstrations and their aftermath, see e.g. Vickers 1998: 197-217.

52 seriously after the rise of Milosevic to the power in the Republic of Serbia in

1987, who started implementing radical policies against the Albanians in

Kosovo, such as the purge of Albanians from the workplaces and the effective annulment of the autonomous status of Kosovo. Against the Milosevic regime, the Albanians organised themselves under the leadership of Ibrahim Rugova, but he chose a policy of peaceful resistance, and thus it did not lead to the armed conflict. However, an underground guerrilla organisation established mainly by the Albanian diaspora in the Western Europe emerged and started its violent activities in 1996. The rebellion escalated into a large-scale conflict in

1998, which culminated in the NATO bombing of Yugoslavia in 1999.

In Macedonia, the Macedonians constituted an absolute majority of the population (66.5% in 1994), while the Albanians constituted a minority

(22.9% in 1994).20 Macedonia had its first democratic elections in November

1990. No party won the majority of the seats in the parliament, which led to the election of Kiro Gligorov as President of Macedonia and the establishment of non-party “expert cabinet” led by Nikola Kljusev, including three Albanian ministers.21 Since the transition to multi-party system in 1990, Macedonia maintained a power-sharing coalition government between Macedonian and

Albanian parties, despite the change of party composition of the government.

Macedonia thus did not see a serious deterioration of the inter-ethnic relations as Serbia did, even though there were several incidents in the 1990s.22

Macedonia also remained calm when it faced a serious challenge posed by the massive influx of the Albanian refugees from Kosovo in 1998 and 1999. In

2001, however, an underground guerrilla organisation appeared in the

20 For the census results in Macedonia from 1961 to 1991, see Woodward 1995: 33. For the census results in 1994, see Peny 1997: 226. 21 As for the results of the elections in 1990, see Perry 1997. 52 For a good overview of the events in the 1990s in Macedonia, see Daskalovski 2005: 51-79.

53 north-western part of Macedonia and started its rebellion activities. The rebellion remained a relatively low-intensity one, however, and was de-escalated after the Ohrid peace agreement in August 2001.

While this thesis treats these four cases separately, this doesnot mean that they were not connected to each other. On the contrary, one can instantly notice a number of linkages between the cases. First of all, the Yugoslav and

Serbian authorities, particularly the Yugoslav military leadership and the

Serbian elite led by Milosevic, were deeply involved in the organisation of the

Serb rebellions in Croatia and Bosnia. Given that the Milosevic regime also played a central role in the escalation of rebellion in Kosovo, one can argue that the same actors, most notably Milosevic and his associates, were involved in three out of the four cases examined here. Indeed, Gow (2003) argues that the primary and defining element in the wars in Croatia, Bosnia and Kosovo was the Serbian state project to create a new equation of territory and inhabitants, i.e. to establish the new borders of a set of territories linked to Serbia, which would be “for” the Serbs and would mostly be ethnically pure.

Secondly, actors in these four cases seem to have learned considerably from the experience of other countries. For example, the organisers of rebellion in Croatia were the politicians from the Serb Democratic Party {Srpska

Demokratska Stranka, SDS), and they were helping their fellows in Bosnia to establish a political party with the same name, the Serb Democratic Party in

Bosnia and Herzegovina (SDS-BiH), which was far more successful in the first elections in 1990 (see Chapter 4). When the SDS-BiH politicians started to prepare for the rebellion, they almost completely imitated what their fellows did in Croatia, who started to rebel a year earlier (see Chapter 4 for the details).

In a similar vein, the organisers of rebellion in Macedonia were veterans of the

Albanian rebellion in Kosovo (see Chapter 4). Indeed, the rebel organisations

54 in Kosovo and Macedonia had the same acronym in Albanian (UQK). The learning process may explain not only the similarity between the cases but also the difference. For example, the Macedonian state authorities took a conciliatory stance towards the Albanian rebels while the Serbian authorities took repressive measures (see Chapter 8 and 9). One may well argue that

Trajkovski, then President of Macedonia, learned from the Serbian experience that the repressive measures can be counterproductive. The international community was also more careful not to use the term “terrorist” in the case of

Macedonia while it denounced the Albanian rebels in Kosovo as “terrorists” many times, which was interpreted as a “green-light” for repression by the

Serbian authorities (see Chapter 9). This difference may also be a product of the learning process.

There are thus a number of linkages between the cases. It would seem only natural: after all, these four countries constituted one country for more than 40 years until 1991 (or more than 70 years if the inter-war period is included). Indeed, as discussed above in this chapter, it has been more common for the scholars to treat the dissolution of Yugoslavia and a series of wars in the region as a single, inter-connected case. This thesis treats the four cases of rebellion in the ex-Yugoslav region separately only to “disaggregate the data adequately” and to study it “as a set of cases,” as proposed by Brubaker and

Laitin (see Section 1 of this chapter). Such a treatment serves to clarify the commonalities and differences between the cases and to build a theory through the comparative method.

As the brief overview of the four cases shows, these four cases present a considerable difference in terms of the initial level of rebellion (between the

Serb rebellions and the Albanian rebellions) and the escalation of the rebellion

(between the Albanian rebellions in Serbia and Macedonia). The present thesis

55 will attempt to explain these differences between the four cases in the comparative case study.

4. Structure of the Thesis: Connection between Theory, Large-N analyses and Case Studies

Finally, this section briefly presents the structure of the rest of the thesis, which is related to the mixed methods approach discussed in this chapter.

The rest of the thesis is divided into three main parts. Each part aims to examine each step of the three-step approach discussed in Chapter 1. In each part, the first chapter (theory and large-N) is followed by chapters) that will conduct an in-depth comparative case-study of the ex-Yugoslav cases on the same subject. The purposes of the case-study are twofold: firstly to conduct a

“controlled comparison” to provide a clear explanation to what happened in the ex-Yugoslav region and to contribute to the better and clearer understanding of these cases; and secondly to evaluate the theoretical argument and the general findings presented in the preceding chapter and to show convincingly that the processes and findings uncovered in the theoretical and statistical analyses have some “real-world” relevance. By adopting such a structure, the present thesis attempts to show a connection between the theory, large-N analyses and in-depth case studies.

The order of the chapters in each part can be a matter of debate. On the one hand, as discussed in Introduction, the present study primarily used the comparative analysis of the ex-Yugoslav cases to develop a theory and then used the large-N analysis to test it. One may thus argue that it would be more natural if the small-N analyses are presented prior to the corresponding large-N analyses. One the other hand, however, the theory and the research design of

56 large-N analysis inform the corresponding case-study as well. For example,

Chapter 6 examines the effect of structural conditions, such as the level of economic development (GDP per capita) and the mountainous terrain, which are used as control variables in the large-N analyses in Chapter 5. A reader will certainly be able to better understand why these factors are examined in the case study if he or she has already read the corresponding theoretical / large-N chapter. In Part IV, Chapter 9 uses two explanatory variables that are not included in the corresponding large-N analyses in Chapter 7, since both of these are extremely difficult to quantify in a convincing manner. Here, the comparative small-N analysis is not used to generate a theory but to compliment the large-N analysis, because variables used in the large-N analysis cannot fully explain the difference between Serbia and Macedonia. In this case, it would be more straightforward to present the large-N analyses prior to the case-study. Due to these merits listed above, it was decided that the theory and the large-N analyses are presented prior to the corresponding small-N analyses. Part II has two chapters. Chapter 3 presents the theoretical discussion inspired by the resource mobilisation theories, and to conduct large-N analyses using the MAR dataset and some variables constructed by the author. Chapter 4 then conducts a comparative case study of four ex-Yugoslav cases, the Serb rebellions in Croatia and Bosnia and the Albanian rebellions in Serbia and

Macedonia.

Part III also has two chapters. Chapter 5 presents theoretical arguments on the “dynamic” grievances and some model to incorporate both dynamic grievances and structural conditions for the explanation of the onset of initial rebellion. It then conducts large-N analyses using the MAR dataset and other variables taken from earlier studies on the onset of civil war. Chapter

6 conducts a comparative case study of the four ex-Yugoslav cases, focusing on

57 the reasons for the rebels to take up arms. It will take into consideration both the objective and structural conditions and dynamic grievances that motivated the rebels to take up arms in these cases.

Part IV consists of three chapters. Chapter 7 presents theoretical arguments on the mechanisms of the escalation of rebellion and the role of state reactions to the initial rebellion. It then conducts a series of large-N analyses to test the hypotheses drawn from the theoretical arguments. Chapters

8 and 9 will conduct in-depth case studies of two ex-Yugoslav cases, namely the Albanian rebellions in Serbia and Macedonia. Because this part examines the escalation of the rebellion from a low-intensity one to a high-intensity one, two cases of Serb rebellions in Croatia and Bosnia are excluded from the case-study, since their initial level of rebellion was very high. Chapter 8 conducts a comparative analysis on the state reactions and their effects on the course of rebellion in Serbia and Macedonia. It will show that the state authorities made different reactions to the initial rebellion, which in turn affected the courses of the rebellion in these two countries. Chapter 9 then conducts an additional comparative analysis on the determinants of the state reactions, attempting to answer the question why the state reactions to the initial rebellion were different in these two countries.

Conclusion

This chapter has laid down the methodological premises for the present thesis. It firstly discussed why the mixed methods approach is adopted in the present thesis. It then discussed the cases for the analysis. Firstly, it discussed the MAR dataset, which will be used as a key dataset in the large-N analyses of the thesis. Secondly, it discussed the four cases of ethnic rebellions

58 in the ex-Yugoslav region, presenting a brief overview of the cases, and why the comparison of these four cases is useful for the purpose of the present thesis. Finally, it briefly presented the structure of the rest of the thesis, which reflects the three-step approach and the mixed methods. Having set the theoretical and methodological premises for the thesis, the thesis will now embark on the empirical analyses in the following chapter.

59 Part II: Initial Level of Rebellion

Chapter 3 Intra-ethnic Division, Decision-Makers and the Rebellion: Quantitative Analysis of the Initial Level of Rebellion

Introduction

Who among the ethnic group actually make a decision to take up arms against the state? This question has not been addressed very much so far, particularly in the game theoretic explanations of ethnic conflict. As discussed in Chapter 1, this is partly a result of the pervasiveness of the unitary actor assumption: when the rebellion occurs, it is assumed that the ethnic group is a cohesive group and somehow reaches the collective decision to take up arms.

But how they do this remains a “black box,” exogenous to the theory being developed. This assumption also methodologically underpins many large-N studies on the occurrence of civil war, which treat the “onset” of civil war as a one-shot event.

However, this unitary actor assumption is problematic, because the ethnic group as such does not necessarily have a clear decision-making system.

This is one of the specific features of domestic ethnic conflict when compared to international war. In the case of international war, the answer to the question

“who makes a decision?” is more obvious: it is the head of the state who makes a final and formal decision to mobilise the national army. International war normally starts with the declaration of war by one country against some other(s), which makes it easy for observers to identify an exact date of commencement of the war. In domestic ethnic conflict, on the other hand, the

60 answer to this question is not necessarily straightforward. Laitin (1998: 331) thus argues that “while it is useful and often powerful to assume that states are unitary actors; it is equally powerful but misleading to assume that nations (or putative nations) are.” Brubaker and Laitin (1998: 438) point out that ethnic conflict differs sharply from interstate conflict, because “[rjarely is a single leader recognized as authoritatively entitled to speak in the name of the group” and thus “ethnic groups generally lack what states ordinarily possess, namely, a leader or leaders capable of negotiating and enforcing settlements.” Fearon

(2004: 407) also argues that the unitary actor assumption is “rarely a plausible assumption” and points out that “instead, nationalist insurgencies are frequently initiated by small minorities within an ethnic group.”

Given this feature of domestic ethnic rebellions, it is important to relax the unitary actor assumption. The present chapter consists of three sections.

The first section presents two ideal types of decision-making within ethnic groups, namely the “decision from above” and the “decision from below,” and their theoretical implications. The second section will present a set of testable hypotheses drawn from the theoretical argument. The following section then conducts a large-N analysis to test these hypotheses.

1. Two Ideal Types: Decisions From Above and Decisions From Below

While a precise answer to the question posed at the beginning of this chapter may vary enormously from one case to another, two ideal types are presented in the present thesis. The first is the decisions made by powerful ethnic leaders, the “decisions from above.” The second is the decisions made by non-leaders within the ethnic group, the “decisions from below.” These two ideal types are based on the vertical intra-ethnic division between leaders and

61 non-leaders.

This vertical intra-ethnic division is important because of the characteristics of domestic ethnic conflict discussed above. As for the interstate war, the decision to go to war can be made only by the political leaders who have competence to do so. On the other hand, in domestic ethnic conflict, anybody in the ethnic group can decide to take up arms against the state, whether they are leaders or non-leaders. One must consider this fact in order to understand the diversity within the category of ethnic rebellion, while few scholars have paid attention to it.

The first type is the decisions made by the ethnic leaders who have significant financial, military, human or other kind of resources that can be utilised for rebellion. Such ethnic leaders can be found in the political and administrative organisations (for example, members of the national parliament, regional governors, or prestigious leaders of the large-scale political movements), military forces (generals, colonels, and so on) and traditional tribes (elderly leaders of tribes, powerful clan leaders, etc). These leaders have either de jure or de facto powers to mobilise resources necessary for the rebellion, such as financial, military, or human resources. These leaders sometimes can make a decision that is binding on a large number, if not all, of members of the ethnic group. If the decision to take up arms is made by these powerful leaders, it is labelled as a decision “from above” in the present thesis.

The second type is the decisions made by non-leaders within the ethnic group who do not possess so significant financial, military or human resources for the rebellion. Such people can be found in the groups of radical students or workers, clandestine or underground radical organisations, self-proclaimed political “party” that does not hold any public office in the state, etc. In such situation, people who make a decision to take up arms are not

62 very powerful in terms of the resources at their disposal. A decision made by such non-leaders may not necessarily be supported by the other members of the ethnic group. It may be opposed and denounced by the ethnic leaders who are prestigious within the ethnic group. These non-leaders might want to replace the incumbent leaders and to take over the political power by taking up arms: indeed, in the post-conflict society, former rebel leaders often become politicians and (try to) replace the old political elite of the ethnic group. In any case, if the decision to take up arms against the state is made by these non-leaders within the ethnic group, it is called a decision “from below.”

These two ideal types correspond to the opposing observations made by different scholars. On the one hand, some have argued for the importance of elites in the occurrence of ethnic violence. For example, by analysing nationalist violence in the former Soviet Union, Beissinger (1998: 418) concluded that the “waves of nationalist violence are disproportionately associated with groups well-connected to the state, not marginalized communities, and violence is more often a strategy for institutionalizing control than for challenging authority.” Mueller has suggested that what is required for the occurrence of ethnic conflict and killings such as those in

Rwanda or the former Yugoslavia is “the breakdown of police order and the mobilization by politicians with their own agenda of small numbers of murderous and opportunistic thugs” (Simons & Mueller 2001: 191, emphasis added). Gagnon (1994) argues that it was a conservative coalition of party leaders, local and regional elites, nationalist intellectuals and segments of the military leadership who provoked violent ethnic confrontation. These observations largely correspond to what is termed a “decision from above”: it is powerful ethnic elites or politicians who tend to lead, organise and realise the violent conflict.

63 On the other hand, other scholars pointed out that the rank-and-file members of the gangs, terrorist groups or guerrilla armies are more often poorly educated and from lower or working class backgrounds, while ethnic leaders are well educated and from middle-class backgrounds.1 In advanced industrial democracies, nationalised violence usually assumes the form of marginalised terrorist activity rather than large-scale mobilisation or interethnic war (Gurr 1993: 98-99). In these cases, the decision to take up arms is not made by the powerful and mainstream ethnic leaders but by a rather marginal segment of the society. These observations correspond to what is termed a

“decision from below.”

Of course, the distinction between elite-driven (“from above”) and grass-roots-driven (“from below”) dynamics itself is far from novel in the study of ethnic conflict. Indeed, many scholars have used this distinction, explicitly or implicitly, in analysing ethnic or internal conflict. For example,

Brown (1997: 15-16) argues that the internal conflicts can be categorised according to whether they are triggered by elite-level or mass-level factors. In analysing the nationalist and regional conflicts in the process of democratisation, Laitin (1995) distinguished “regional elite” and “vigilantes” within the minority group as distinct players in the four-player game (“state” and “people” constitute the other two players) and argued that “policing the vigilantes” who might resort to terrorism is a key concern in the making of credible promises by regional leaders. In two separate monographs, Petersen addressed the puzzle of “how the less powerful manage to sustain rebellion against the more powerful” (Petersen 2001) and the question of “why and when the more powerful commit violence and discrimination against the less powerful” (Petersen 2002). Bourne (2007) made a distinction between the

1 See, for example, Clark 1984. It is cited in Brubaker & Laitin 1998: 434.

64 top-down mode and the bottom-up mode in the arming process of warring

factions in the internal conflict. These authors distinguish elite-driven from

grass-roots-driven dynamics, even though definitions of “elite” (or “the

powerful”) and “grass-roots” (“mass”, “vigilantes” or “the less powerful”)

differ from author to author.

The novelty of the present chapter does not lie in the distinction

between two ideal types per se, but lies in connecting it to an under-researched

aspect of the ethnic conflict, namely the initial level of rebellion. As discussed

in Chapter 1, the initial level of rebellion has rarely been systematically studied

in the existing literature, partly due to the unitaiy actor assumption and the

conceptualisation of the “onset” of large-scale rebellion as a one-shot event:

when one assumes that the ethnic group decides to take up arms en masse, and the occurrence of large-scale rebellion is a one-shot event, the initial level of

rebellion will be always high by assumption. In reality, however, the initial

level of rebellion made by members of the ethnic group is not always high. Indeed, as the large-N analysis of this chapter will show, the high level of

initial intensity occurs relatively rarely. If the initial intensity of rebellion varies

from case to case, a theory of ethnic rebellion must explain why it becomes

high in some cases while it becomes low in other cases.

How is this distinction between the two ideal types connected to the

initial intensity of rebellion? This chapter hypothesises that the initial rebellion

tends to be higher when a rebellion is directed by the decisions from above,

while it tends to be lower when a rebellion is directed by the decisions from below. The causal factor that explains this connection is the amount of resources available to the decision-makers. As the resource-mobilisation theorists emphasised, resources are of particular importance for collective action: recruitment and logistics for collective action require a considerable

65 amount of time, money and effort, and various kinds of resources such as financial, technological, human resources must be utilised (e.g. Oberschall

1973; McCarthy & Zald 1977; Jenkins 1983). The resources available to the organisers of the collective action, therefore, will affect the course and the outcome of the action.

This logic of the resource-mobilisation theory is applicable to ethnic rebellion. If the rebellion is directed by a “decision from above,” those who make a decision to take up arms possess significant financial, military or human resources for the rebellion. As a result, their decision to take up arms is more likely to lead to the occurrence of large-scale rebellion. After such an occurrence, the state authorities are practically drawn into the civil war whether they want to or not, and they are forced to take all necessary measures to fight against the rebels and to maintain order in the country. If the rebellion is due to a “decision from below,” those who make a decision to take up arms do not possess sufficient resources for large-scale rebellion. As a result, it is more likely that their decision to take up arms will lead to the occurrence of rebellion of a relatively low-level intensity, and would not necessarily lead immediately to the occurrence of the large-scale rebellion. This argument can be summarised as presented in Table 3.1.

Table 3.1 Type of Decision-makers and the Initial Level of Rebellion

Type of decision-makers Initial level of the rebellion

Powerful ethnic leaders High (Decision from above)

Non-leaders Low (Decision from below)

66 Needless to say, this argument does not mean that this intra-ethnic factor is the sole determinant of the initial level of rebellion. There are other factors that may also affect the initial level of rebellion. Among them, inter-ethnic factors deserve special attention because of their direct relevance to the intra-ethnic factor. The position and status of the ethnic groups within the state and their relationship to the state authorities vary considerably across ethnic groups. Some ethnic groups may have easy access to a large amount of resources while others may not. Some ethnic group may be in such unfavourable conditions that they cannot have powerful political elite at all.

The status of the ethnic group vis-a-vis other ethnic groups within the state may limit the amount of resources available to the ethnic leaders, and thus the initial level of rebellion may differ across ethnic groups even when the rebellion is directed by the ethnic leaders. Table 3.2 summarises this inter-ethnic factor. By combining these two arguments above, one may make a

2 by 2 table, which is presented in Table 3.3.

Table 3.2 Status of the Ethnic Group within the State and the Initial

Level of Rebellion

Status of the Ethnic Group Initial level of the rebellion

Powerful Ethnic Group High

Weak Ethnic Group Low

67 Table 3.3 Type of Decision-makers, the Status of the Ethnic Group and the Initial Level of Rebellion

Powerful Ethnic Group Weak Ethnic Group

Leaders Initial Level: Very High Initial Level: Relatively High

Non-leaders Initial Level: Relatively High Initial Level: Low

Of course, this 2 by 2 table cannot be so precise: for example, it cannot specify whether the initial level of rebellion led by non-leaders of powerful ethnic groups will be higher or lower than that of rebellion directed by leaders of weak ethnic groups. By making such a table, however, one can explicitly differentiate the intra-ethnic and inter-ethnic factors and avoid confusing these two factors.

In any case, the intra-ethnic factor discussed above is of the key interest of the present chapter because of its relevance to the critique of the unitary actor assumption. In order to evaluate the validity of the argument, the following sections will attempt to make testable hypotheses and to conduct large-N analyses to test them.

2. Drawing Testable Hypotheses

This section will draw a set of testable hypotheses based on the argument developed above. To recapitulate, it was argued that a rebellion is more likely to be large-scale at the outset if the powerful ethnic leaders decide to take up arms, while it is more likely to be small-scale if the non-leaders decide to do so. The key task in making testable hypotheses based on this

68 argument is to draw a clear line between the “powerful ethnic leaders” and

“non-leaders.” One must be careful not to make vague criteria for drawing this

line, especially because of the possibility that knowing the results can affect the judgment. If criteria for drawing the line are vague, one may categorise the

decision-makers as “powerful elite” precisely because the initial level of

rebellion was high, and may consider them to be “non-leader” because the

initial level of rebellion was low. In this case, one will surely observe the

connection between the type of decision maker and the initial level of rebellion,

but it is not because the theory is true but simply because the judgment on the

type of decision maker is affected by the knowledge of the results. In order to

avoid such a mistake, criteria for drawing line should be clear and totally independent from the response variable.

This chapter uses the findings in the comparative case-study, which

will be presented in the next chapter, to draw testable hypotheses with clear

criteria for drawing the line between leaders and non-leaders. Therefore, while the hypotheses and the large-N analyses are presented here before the

comparative case-study, these hypotheses and the research design of the

large-N analyses result from the feedback from the qualitative case-study. In this chapter, two testable hypotheses are drawn from the comparative

case-study.

Firstly, one specific feature of the Serb rebellions in Croatia and

Bosnia was that the rebellion was planned, organised and realised by the politicians who held public offices in the countries. In the case of Croatian

Serbs, the important figures who led the rebellion held public offices at the

local level, such as Milan Babi6 (mayor of Knin municipality) and Milan

Martic (an inspector of the police force in Knin). The leading figures among the Bosnian Serbs held public offices at the republic level, such as Radovan

69 Karadzic (a member of the Bosnian national parliament), Biljana PlavSic (a member of the seven-member presidency of Bosnia), Momcilo Krajisnik

(president of the Bosnian national parliament), and the SDS also controlled a significant share of the national and local institutions. In both cases, the SDS politicians could use a significant amount of resources to prepare, organise and realise the rebellion. If one generalises this observation on the Serb rebellions, one can expect that the initial level of intensity of the rebellion will be higher if the rebellion is led by the politicians who hold public office in the state, because it is highly likely that these politicians can use various kinds of resources to plan and organise a large-scale rebellion. One can thus draw the following hypothesis:

Hypothesis 1: If the rebellion is directly led by the politicians who hold public

office, whether at national or local level, it is more likely that the initial

level of intensity of rebellion would be higher.

Another feature of the Serb rebellions was that members of the local police force and a part of the regular national army took sides with the rebels and directly participated in the rebellion. The police force and the national army are the key institutions for the monopoly of violence within the state. The participation of these two institutions in the rebellion thus means not only that the rebels have more military and human resources for the rebellion but also that there is less policing against the rebel activities. In the case of Serbs in

Croatia and Bosnia, the participation of the police force and the national army with a large amount of heavy weapons led to an occurrence of rebellion of high intensity, while the amount and the size of the weapons available to organisers of the rebellions were much more limited the in case of Albanians in Serbia and

70 Macedonia, where no police or regular national army participated in the rebellion. By generalising this observation, therefore, one can expect that the initial level of intensity of the rebellion would be higher if members of police force or the regular national army participate in the rebellion from the outset of the rebellion. Thus:

Hypothesis 2: If members of the police force or the national army participate in

the rebellion from the outset, the initial level of intensity of rebellion

would be higher.

Of course, politicians who hold public office and members of the police force / national army are not the only type of ethnic leaders that would make a decision to take up arms. In many cases, ethnic leaders that make a decision to rebel include clannish/tribal leaders (who do not hold any public office) or leaders of the large militia that do not belong to the national army.

The two hypotheses presented here cannot capture these types of leaders and thus they would only explain a part, not the whole, of the rebellions of high initial intensity. The advantage of these two hypotheses is, however, that criteria are clear and judgment would not be affected by the knowledge of the results. For example, whether the initial level of rebellion was high or not cannot affect the judgment on whether leaders who led the rebellion held public office at that time or not. On the contrary, when it comes to clan/tribe or militia, the definition of “clan” or “militia” could be diverse and judgment could be more subjective. By excluding these ambiguous cases, therefore, one could avoid the danger of being affected by the knowledge of the response variable in drawing lines between leaders and non-leaders.

As discussed above, the intra-ethnic factor is not the sole factor that

71 would affect the initial intensity of rebellion. One important control variable that should be taken into consideration is the inter-ethnic factor. In addition, one can derive more control variables from some key findings in the existing literature on ethnic conflict and civil war. In order to avoid the danger of the omitted variable bias, one must take into consideration these control variables that would also affect the initial level of rebellion.

It is extremely difficult to measure quantitatively the resources available to the ethnic group within the state. One indicator that can be used - and relatively easily measured quantitatively - is the relative size of the ethnic group within the state. For a number of reasons, one can argue that the relative size of the ethnic group would affect the resources available for rebellion.

Firstly, the relative size of the ethnic group will primarily determine the maximum human resources available for the rebellion: if the group is relatively small, it would be harder to recruit sufficient number of fighters against the state. Secondly, the relative size of the ethnic group will determine the balance between that ethnic group and other group(s) that control the state authorities.

As the relative size of the ethnic group gets larger, the size of this group and that of other group(s) controlling the state authorities will get closer and thus the state authorities may not be so powerful vis-a-vis that ethnic group. These arguments suggest that it is more likely that the initial level of rebellion would be higher as the relative size of the ethnic group gets higher.

In addition to the inter-ethnic factor, some more control variables will be included in the analysis below. In the present analysis, the following variables are included as control variables: (1) the level of economic development, (2) political regime of the country, (3) newness of the state, (4) political instability, and (5) the terrain of the country. These control variables are derived from some key findings of earlier research.

72 According to some large-N analyses, the level of economic development is one of the strongest predictors of the onset of “civil war”

(Fearon & Laitin 2003; Collier & Hoeffler 1998). These scholars have identified several mechanisms that can be behind the strong association between low levels of economic development and the onset of civil war. For example, Collier and Hoeffler (1998: 565-567) argue that this factor represents the “opportunity cost” of rebellion for individuals: the higher economic development is, the higher the opportunity cost is for individuals to stop economic activities (i.e. give up earnings from normal economic activities) and to take up arms. Fearon and Laitin (2003: 80) argued that GDP per capita can be a proxy for three mechanisms: (1) overall financial, administrative, police, and military capabilities of the state, (2) the degree of penetration into rural society by the central administration, (3) the possibility of recruiting young men to the life of guerrilla. Unfortunately, as Humphreys (2005: 533) pointed out on the relations between natural resources and civil war, the scholars have

“focused on correlations without constructing tests to identify particular mechanisms that may underlie those correlations” and has “arbitrarily favored one mechanism to the exclusion of others” and it is still not clear which mechanism, or a combination of mechanisms, is working behind the association between the level of economic development and the onset of civil war. Whichever mechanism is working, however, one can assume that this variable would also affect the initial level of rebellion by the same logic. If the low level of economic development means low opportunity cost for potential rebels, it would be easier for organisers of rebellion to recruit more fighters if the level of economic development is low, and thus the initial level of rebellion would be higher. If the low level of economic development means weaker police and military capabilities of the state, it would be easier to organise a

73 rebellion of a higher level of intensity if the level of economic development is low. In any case, therefore, one would expect that it is more likely that the initial level of rebellion will be higher as economic development gets lower.

Secondly, the political regime of the country would also affect the initial level of rebellion. If the political regime in the country is democratic, there are some institutional channels for the expressions of individual and collective preferences and greater prospect of peaceful change of the status quo.

While it may not always prevent the emergence of a small number of extreme radicals, the presence of a democratic political regime in the country would make it harder for the organisers of the rebellion to persuade other members among the ethnic group to take up arms against the state. On the contrary, if the political regime is undemocratic, the prospect for the peaceful change of the status quo is much bleaker. It would thus be easier for the organisers of the rebellion to persuade other people to take up arms with them. Gurr (2000), for example, finds that the presence of democratic institutions is associated with peaceful demonstrations while the authoritarian regime is associated with violent rebellions. Thus, one would expect that it is less likely that the initial level of rebellion would be higher as the political regime gets more democratic.

Finally, three factors associated with the “right conditions of insurgency” identified by Fearon and Laitin will be used here, namely the newness of the state, political instability, and the geographic terrain of the country. Firstly, Fearon and Laitin (2003: 81) argue that it is easier for the potential rebels to organise a rebellion in a newly independent state, since it often loses suddenly the coercive backing of the former imperial power and its military capabilities are new and untested. The same logic can be applied to the explanation of the initial level of rebellion. Technically, it would be easier for the potential rebel leaders to organise a large-scale rebellion if the police and

74 military capabilities are not established in the country. In addition, it would be easier to persuade members of the ethnic group to take up arms if the police and military capabilities are not established, since it is less likely that they would face harsh punishment and retaliation by the police and military of the state authorities. Politically, it would be easier for the organisers to persuade people to take up arms if the legitimacy of the new state is not yet established.

Secondly, by the similar logic, political instability is also considered to favour insurgency, since the political instability at the centre indicates disorganisation and weakness of the central authorities and thus an opportunity for a rebellion

(Fearon and Laitin 2003: 81). Thirdly, Fearon and Laitin argue that a rough terrain would also favour insurgency since it would be easier for the rebels to hide from the government forces and to sustain the insurgency. By the same logic, it would be easier to organise a rebellion if the terrain is rough, since the rough terrain would deter the governments from effectively policing potential rebels. Their findings suggest that it is more likely that the initial level of rebellion would be higher (1) if a country is a newly independent state, (2) if a country is experiencing political instability at the centre, or (3) in a country where the terrain is rougher.

3. Testing Hypotheses: Large-N Analyses

This section will conduct large-N analyses to test the hypotheses drawn in the previous section. Firstly, it will conduct a bi-variate analysis to examine the relationship between the intra-ethnic factor and the initial level of rebellion. It will then conduct a multi-variate analysis to consider the effect of the control variables.

The MAR dataset is used for the case selection. The MAR dataset

75 contains ordinal data on conflict behaviour of more than 300 ethnic groups from 1945 to 2000 (for five-year period from 1945 to 1984 and for each year from 1985 to 2000). For the present analysis, an ordinal measure of

“anti-regime rebellion” will be used, which is coded with the scale presented in

Appendix 1. This variable enables us to identify (1) which ethnic groups were engaged in rebellion activities and (2) what was the initial level of the intensity of rebellion.

From the MAR dataset, the author identified all ethnic groups that had been engaged in some level of rebellion during the period from 1945 to 2000, and checked what was the level of intensity of the rebellion when they started being involved in the rebellion activities for the first time during this period (i.e. the initial level of intensity of rebellion). Out of more than 300 ethnic groups included in the MAR dataset, 177 ethnic groups were engaged in rebellion at some time during the period, and the initial level of rebellion of these ethnic groups varied from 1 (sporadic terrorism) to 7 (protracted civil war). These 177 ethnic groups are the cases in the large-N analyses conducted below (the unit of analysis is the ethnic group).

(1) Bi-variate analysis

This section firstly conducts bi-variate analysis to test the two key hypotheses discussed above (HI and H2). In order to do so, one must know (1) whether these rebellions were led by the politicians who held public office at the time of commencement of the rebellion, and (2) whether members of the police force or the national military participated in the rebellion when the rebellion started. While the MAR dataset contains more than 400 variables, there is unfortunately no variable that enables us to check these two points for the cases of rebellion. The present study, therefore, surveyed (1) what was the

76 organisation behind the rebellion at the initial stage, (2) whether the

organisation was led by the politicians who held some kind of public office,

and (3) whether members of the police force or national military participated in

the rebellion at the initial stage, for all 177 ethnic groups. The resources used

for the judgment are primarily the MAR qualitative data2 but also various

other secondary resources, such as the human rights reports of the US

Department of State, the reports published by prominent NGOs like Amnesty

International and Human Rights Watch, academic articles, journalistic accounts

on the events, interviews of the leaders of the rebels, and in some cases, the

website of the organisations that were behind the rebellion, since the MAR

qualitative data often lacks specific information on the organisation that was

behind the rebellion when it started. The results of this survey are presented in

Appendix 7. Based on the results of the survey, two dummy variables are

created. The first is the POLTCIAN, which takes a value of 1 if the

organisation that started the rebellion was led by the politicians who held some

kind of public office. The second is the POLARM, which takes a value of 1 if

members of the police or national army participated in the rebellion when it

started.

To test the first two hypotheses of our theoretical interest (HI and H2),

a bi-variate analysis will be conducted by making a cross-tabulation. If

hypotheses above are right, a higher level of rebellion must be more frequent

when the dummy variables take the value of 1. The Pearson's chi-square

statistic will be used to judge whether there is any association between the two

variables.

2 MAR qualitative data contains a brief summary of the history and the situation of each ethnic group as well as the chronology of key events related to the ethnic group. The data can be obtained via the website of the MAR project: http://www.cidcm.umd.edu/inscr/ mar/index.htm.

77 Let us firstly examine the relations between the initial level of rebellion and the engagement of politicians in the rebellion. The cross-tabulation of the two variables is presented in Table 3.4. This table supports hypothesis 1 discussed above: while the lowest intensity of rebellion

(REB=1) constitutes 34 percent of all cases when politicians are not involved in the rebellion (POLTCIAN = 0), there is absolutely no case of the lowest intensity of rebellion when politicians are involved (POLTCIAN =1); while the largest intensity of rebellion (REB=7) constitutes only 8 percent of all cases when politicians are not involved in the rebellion, it constitutes 50 percent of all cases when politicians are involved. One can argue therefore that the initial level of rebellion is skewed toward the higher level when politicians are involved in the rebellion. The Pearson’s chi-square statistic for this table is 34.6, statistically significant at the 0.01 level. Therefore, one can reject the null hypothesis that there is no association between these two variables.

Table 3.4 Initial Level of Rebellion and Involvement of Politicians, 177 ethnic groups from 1945 to 2000

Initial Level Involvement of the Politicians0 Total of Rebellion No (0) Yes (1) 54 0 1 54 (34.4%) (0 %) 16 18 ( 1 0 .2 %) ( 1 0 .0 %) 25 4 3 29 (15.9%) (2 0 .0 %) 25 1 4 26 (15.9° o) (5.0%) 4 1 5 5 (2.5° o) (5.0%) 21 -> 6 23 (13.4%) ( 1 0 .0 %) 12 10 7 (7.6%) (50.0%) Total 157 2 0 177 Pearson chi2(6) = 34.6 P < 0.001

78 It should be noted that the involvement of politicians is observed in a relatively small number of cases of rebellion in the MAR dataset: in total, it numbers only 20 cases (11%) out of 177 ethnic rebellions observed in the period. This shows that, generally speaking, the involvement of politicians in the organisation of ethnic rebellion is relatively rare. It seems that there are two major reasons for this. Firstly, in some cases, ethnic groups are so disadvantaged in the country that they lack political elites who can hold some kind of public office in the country at all. This seems the case, for example, for

African-Americans in the USA until the 1960s, some (but not all) indigenous peoples in Southern America, and some ethnic minorities under harsh dictatorships in Sub-Saharan Africa. Secondly, in other cases, ethnic groups did have political elites who held some kind of public office, but they were not involved in the ethnic rebellion because they were more moderate than those who actually took up arms. The qualitative analysis in the following chapter will show, for example, that the Albanians in Kosovo and Macedonia fall into this category. In the case of Sri Lanka, Tamil political parties held some seats in the Sri Lankan parliament, but these Tamil political elites were against violence and did not organise the rebellion.

The fact that the involvement of politicians in the organisation of ethnic rebellion is relatively rare naturally raises one question: if it is rare, does this variable really explain the variance observed in the response variable? An answer to this question will be explored in the multi-variate analysis conducted below.

Let us now examine the relations between the initial level of rebellion and the engagement of the police force or national army in the rebellion. The cross-tabulation of the two variables is presented in Table 3.5. The association

3 For a detailed analysis on the case of Sri Lankan Tamils, see e.g. Bose 1994.

79 between the two variables is weaker compared to Table 3.4, but the initial level of rebellion is still skewed toward the higher level when members of the police or army were involved in the rebellion: while the lowest intensity of rebellion

(REB=1) constitutes 32 percent of all cases when members of police or army are not involved in the rebellion (POLARM = 0), it constitutes 20 percent if they are involved in the rebellion (POLARM =1); while the largest intensity of rebellion (REB=7) constitutes 9.6% of all cases when members of police or army are not involved in the rebellion, it constitutes 35% of all cases when they are involved. The Pearson’s chi-square statistic for this table is 14.4, which is much smaller than that for Table 3.4, but is still statistically significant at the

0.05 level. Therefore, one can reject the null hypothesis that there is no association between these two variables.

Table 3.5 Initial Level of Rebellion and Involvement of Police or Army, 177 ethnic groups from 1945 to 2000 Initial Level Involvement of the Police or Army? Total of Rebellion No (0) Yes (1) 50 4 1 54 (31.8%) (2 0 .0 %) 18 0 o 18 (11.5%) (0 %) 25 4 3 29 (15.9%) (2 0 .0 %) 25 1 4 26 (15.9%) (5.0%) 4 1 C 5 (2.5%) (5.0%) 2 0 3 6 23 (12.7%) (15.0%) 15 / i (9.6%) (35.0%) Total 157 2 0 177 Pearson chi2(6) = 14.4 P < 0.05

80 Why is the level of rebellion low even when members of the police or national army participate in the rebellion? A closer look into the cases suggests the importance of the number of those who participated in the rebellion.

Among the cases where the POLARM is coded 1, the low-intensity rebellion is coded primarily when the actual coup or coup-attempt occurred and thus only a few soldiers participated in the activity. Even when they have some weapons at their disposal, their participation cannot lead to a large-scale rebellion if the number of participants is so small. On the contrary, the high-intensity rebellion is coded when some units or groups of soldiers participate in the rebellion en masse, as the police force and the JNA did for the Serb rebellion in the case of ex-Yugoslavia. This suggest that finer coding is necessary to test hypothesis 2 presented above, which distinguishes the participation of only a few individuals that belong to the police or military and the participation of units or groups of those who have some weapons. This, however, raises a number of practical problems. Firstly, it is difficult to establish an objective and convincing criteria for the distinction between a “group” and “a few” individuals - the criteria could be 10, 20, 50, 100, 200, or higher number of individuals and it is difficult to judge which is adequate and which is not.

Secondly, it is extremely difficult to acquire information on exactly how many individuals of the police or military actually took part in a rebellion. For these reasons, further disaggregation is not conducted and the variable of POLARM is used in the multi-variate analysis despite these reservations.

(2) Multi-variate analysis

Bi-variate analyses conducted above appear to support the two key hypotheses of this chapter (HI and H2). In order to adequately evaluate the explanatory power of the key variables of interest, however, one must conduct

81 a multi-variate analysis that includes control variables that may affect the initial level of rebellion. The variables of the theoretical interest here are POLTCIAN and POLARM discussed above. If the two key hypotheses of this chapter are true, these variables should be associated with the higher level of initial rebellion even when other control variables are controlled for. This part will firstly present the operationalisation of the control variables, secondly the estimation method, and thirdly the empirical results and interpretations.

As for the relative size of ethnic groups within the state, a variable

GPRO in the MAR dataset will be used. This is a continuous variable, ranging from 0 to 1, which shows the proportion of the estimated group population in the country population in the 1990s. This variable thus captures the relative size of ethnic groups within the state. The drawback of this variable is that the data is available only for the 1990s. If one simply plugs in this variable in the dataset, therefore, data will be missing for 137 ethnic groups (77 percent) out of total 177, which will be dropped from the analysis when one conducts the large-N analysis. It is obvious that one cannot evaluate the explanatory power of the variables adequately if more than three-fourths of total cases are dropped from the analysis. In order to avoid this, the average value of GPRO for the period when data is available is applied to the whole period of the dataset in the present analysis (this variable will be called GPROAV hereafter) and is used in the present analysis. This manipulation of the data is of course not ideal, especially if the population of the ethnic group changes drastically during the period when the GPRO data is not available. Given the lack of comprehensive data, however, this is the only way to control for the relative size of the ethnic group within the state, and GPROAV is used below despite these drawbacks.

As for the other control variables, the present analysis will follow the operationalisations of earlier research. As a proxy of the level of economic

82 development, it uses GDP per capita of the country in which the ethnic group live, taken from the Penn World Tables, following the large-N analyses conducted by Collier & Hoeffler and Fearon & Laitin mentioned above (this variable will be called GDP hereafter). As a proxy of the degree of the democratic nature of the political regime, the “polity2” variable in the PolityIV dataset will be used (POLITY2 hereafter). This variable takes the value ranging from -10 (strongly autocratic) to +10 (strongly democratic)4 and is used widely for large-N analyses as a proxy of the degree of democracy of the political regime in the country. As for the newness of the state and the instability in the centre, the present analysis uses dummy variables created by

Fearon and Laitin (2003): the new state dummy (NEWST hereafter) takes the value of 1 in the first and second year of independence of the state, and the political instability dummy (INSTAB hereafter) takes the value of 1 when the country had a three-or-greater change on the Polity IV regime index in any of the three years prior to the year in question. These four variables change their values over time, and the present analysis will use their values at the time of occurrence o f initial rebellion. As for the rough terrain, the variable of

“mountainous terrain” used by Fearon and Laitin is used in the present analysis

(MONT hereafter).5 The response variable is naturally the initial level of the rebellion (REB hereafter), which takes a value from 1 to 7.

As for the estimation method, the ordered logit model is used because of the ordinal nature of the response variable. While some people use the

Ordinary Least Square regression for the ordinal response variable, as if it is a continuous interval variable, it is better to use models specifically designed for

4 For detailed description of the variable and the coding procedure, see the “dataset user’s manual of the Polity IV project,” available online at: http://www.cidcm.umd.edu/ inscr/politv 5 This variable is based on the codings of geographer A.J. Gerard and was expanded by Fearon and Laitin. See Fearon & Laitin 2003: 81.

83 ordinal variables, given the risk that the OLS regression of an ordinal outcome provides misleading results (Long 1997: 115). Using the most appropriate estimation method will reduce the risk of drawing a wrong conclusion.

Based on the construction of the ordered logit model as a latent variable model,6 the fitted model of the present analysis can be summarised as follows:

1 1 Pr (REB, = m | x) = ------1 . (-rm + xi) 1 , (-Tm-j+xb) l + exp l + exp where Pr (REB/ = m) = probability that the level of initial rebellion by ethnic

group / takes the value of m (m = l to 7)

xj3 = (/^POLTCIAN/ + /?2POLARM/ + /^GPROAVz + faGDFi +

J&POLITY2/ + ANEWST/ + #INSTAB/ + /W ONT/)

The cut points7 and the coefficients for the independent variables are estimated by STATAver. 8.

Results of the ordered logit analysis are presented in Table 3.6. Among the eight independent variables, the effect of five variables is statistically significant at least at the 0.1 level, while the effect of the variable POLARM,

NEWST and GPROAV is not statistically significant. This suggests that, at least based on the dataset analysed here and the coding for the participation of members of police and army employed in the dataset, there is no statistical

6 For the explanation of the ordered logit model as a latent variable model, see Long & Freese 2003: 151-155; Agresti 2002: 277-279. 7 The cut points are represented by r in the formula. For J ordinal categories, it is assumed that T0 - -o o , Tj = oo, and y< = m if rm_/ < (x , |3+ £ ,) < ^ for m = l to J. See Long & Freese 2003: 151-155; Agresti 2002: 277-279.

84 evidence that the participation of members of police or army has a significant

effect on the initial level of rebellion. As discussed above, this result for the

POLARM seems to be partly due to the coding procedure that treats the

participation of a few members in coup attempts and the participation of groups

or units of police and army in the rebellion as equal. However, since it is

difficult to disaggregate the data further in a convincing manner, the effect of

the participation of police and army is not explored further here.

Table 3.6 Ordered Logit Analysis of the Initial Level of Rebellion, 177 Ethnic Groups ______Variables Coef. Std. Eit. Involvement of Politicians (POLTCIAN) 1 9 4 *** 0.57 Involvement of Police/Army (POLARM) 0.33 0.51 Group Proportion in Population (GPROAV) -1.40 087 Economic Development (GDP) -0.16** 0.06 Political Regime (POLITY2) -0.09*** 0 0 2 New States (NEWST) 0.76 0.50 Political Instability (INSTAB) 0.71* 0.40

Mountainous Terrain (MONT) 0.30** 0 12 Cut Point (1) -0.40 0 4 4 Cut Point (2) 0.19 0.44 Cut Point (3) 1.07 0.44 Cut Point (4) 1.91 0.46

Cut Point (5) 2.11 0 4 7

Cut Point (6 ) 3.25 0.51 Number of obs = 165; LR clii2(7) = 58.13, Prob > chi2 = 0.01 * p < 0.1; ** p < 0.05, ***p<0.01.

Note, however, that the other variable of theoretical interest, namely

POLTCIAN, is statistically highly significant at the 0.01 level, even when other variables are controlled for. The sign of the coefficient is positive and thus supports hypothesis 1 presented above: the participation of politicians in the organisation of rebellion increases the probability that the initial level of

85 rebellion would take a higher value, other conditions being equal. Note that the

effects of all control variables that are statistically significant are in accordance

with the theoretical expectations stated above. The signs of the coefficients of

GDP and POLITY2 are negative, which means that the probability that the

initial level of rebellion would take a higher value decreases as the level of

economic development increases and the political regime of the country

becomes more democratic, other conditions being equal. The signs of the

coefficients of INSTAB and MTNEST are positive, which means that the

probability that the initial level of rebellion would take a higher value increases

if a country is experiencing political instability, and as the terrain of the country

becomes more mountainous. Based on these empirical results, one can reject

the null hypothesis that the participation of the politicians does not have any

effect on the initial level of rebellion.

Going beyond sign and significance, one can estimate the magnitude

of the effect of our key variable. Firstly, one can calculate the marginal change in the predicted probability and see what the effect of the change in the values

of the variable is. As for the variable POLTCIAN, thechange in the predicted

probability is -0.28 for the value 1 of the rebellion, -0.09 for the value 2, -0.08

for the value 3, 0.05 for the value 4, 0.03 for the value 5, 0.18 for the value 6,

0.20 for the value 7, holding four continuous variables (GPROAV, GDP,

POLITY2, MONT) at the mean and assuming that the other dichotomous variables (POLARM, NWSTATE, INSTAB) takes the value of 0.8 These

numbers indicate how much the predicted probability changes when the value of the variable POLTCIAN changes from 0 to 1 while other variables are held constant. For example, the change in the predicted probability is 0.20 for the

o These values are calculated by STATA ver. 8 , using prchange command. For the details of the procedure, see Long & Freese 2003: 178-181.

86 value of 7. This means that the predicted probability that the initial level of rebellion takes the level 7 increases by 20 percent when the value of variable

POLTCIAN changes from 0 to 1 (i.e., when the politicians are involved in the

rebellion), holding other variables at the value stated above. When the change

in predicted probability takes negative value, this means that the predicted probability will decrease. For example, the change in the predicted probability is -0.28 for the value 1 of the rebellion. This means that the predicted probability that the initial level of rebellion takes the level 1 will decrease by

28 percent when the value of variable POLTCIAN changes from 0 to 1, holding other variables at the value stated above.

Secondly, one can also calculate the change of the odds in order to see the effect of the variables.9 As for the variable POLTCIAN, the results presented in Table 3.6 mean that the odds of having a higher level of rebellion become seven times higher if the politicians are involved in the rebellion, holding all other variables constant. These post-estimation analyses show that the effect of the variable POLTCIAN is fairly strong compared to the effect of other variables. For example, the odds of having a higher level of rebellion become only twice higher if the value of the variable INSTAB changes from 0 to 1 (i.e. when the country experiences political instability). One can conclude, therefore, that the multi-variate analysis strongly supports hypothesis 1 of the present chapter.

9 “Odds” can be calculated as follows:p / (1 -p), where p is the probability of occurrence. If the odds take a value of 1, this means thatp is equal to 0.5, meaning that the chance of the occurrence and non-occurrence is even-even. If the odds takes a value larger than 1, this means that p is larger than 0.5, meaning that the chance of occurrence is larger than non-occurrence. On the contrary, if the odds take a value less than 1, this means thatp is smaller than 0.5, meaning that the chance of occurrence is smaller than non-occurrence.

87 Conclusion

This chapter argued for the importance of intra-ethnic factor in explaining the variance in the initial level of rebellion. It firstly presented two ideal types, (1) rebellion from above, led by powerful ethnic leaders, and (2) rebellion from below, directed by non-leaders. After a brief discussion of their relationship with the inter-ethnic factor, this chapter made two key hypotheses from the theoretical argument. It then conducted large-N analyses using the

MAR dataset and some dummy variables constructed by the author specifically for the present analysis. This chapter has found that, while the second hypothesis was not supported by the statistical analysis, results of both bi-variate and multi-variate analyses support the first hypothesis. When politicians who hold the public office participate in the organisation of ethnic rebellion, the initial level of rebellion is more likely to be higher than otherwise. The theoretical and methodological implications of the findings in this chapter will be discussed in the concluding chapter. In the following chapter, the theoretical arguments of the present chapter will be applied to the cases of ex-Yugoslav region to examine the causal mechanisms in the real-life situations.

88 Chapter 4 Decisions from Above and from Below: Comparative Analysis of the Serb and Albanian Rebellions

Introduction

This chapter conducts a case study of the four cases of rebellions in the ex-Yugoslav countries, namely in Croatia, Bosnia, Serbia (Kosovo) and

Macedonia. It aims primarily to show that the theoretical arguments made in the previous chapter are based on some real cases and to make generalised arguments empirically more convincing. Chapter 3 argued that a rebellion is more likely to be large-scale at the outset if the ethnic leaders decide to take up arms, while it is more likely to be small-scale if the non-leaders decide to do so. In line with that argument, this chapter will show that the level of initial intensity of the rebellion was high when the ethnic leaders decided to rebel (the case of Serbs in Croatia and

Bosnia) while it was low when non-leaders decided to take up arms (the case of

Albanians in Kosovo and Macedonia). The key factor that connects the type of decision-maker with the level of initial intensity of rebellion is the amount of resources available to the decision-maker. When the leaders make a decision to go to rebellion, they can mobilise a large amount of resources and thus can make a large-scale rebellion from the outset. When non-leaders make a decision to go to rebellion, the amount of resources they can mobilise is far more limited, and they could only make a relatively small-scale rebellion at the beginning.

This comparison of the four cases can make an important contribution in the field of ex-Yugoslav studies, because scholars tend to put more emphasis

89 on the top-down nature of the conflict in this region based on the findings mainly from Serbia, Croatia and Bosnia. As discussed in Chapter 2 (p. 54), Gow

(2003) insists upon the Serbian state project as the primary and defining element in the war in this region. While he analyses the case of Kosovo, he only examines the actions made by the Serbian authorities without making equal efforts to analyse and explain the actions made by the Albanian rebels.

Gagnon (2004) argues that the war in the ex-Yugoslav region was caused by the conservative elite such as Milosevic in Serbia and Tudjman in Croatia who wanted to demobilise the opposition against the regime. In his reply to the critique of his book, he argued that “the violence did not happen in a vacuum, nor was it a spontaneous eruption from below. It was the result of conscious strategic planning from above” (Gagnon 2007: 217). His analysis, however, does not include the case of Kosovo or Macedonia. Given this lack of equal attention to the cases of Kosovo and Macedonia, it would be legitimate to question whether the arguments in the existing literature are generalisable even within the region o f ex-Yugoslavia.

Gagnon does not seem to be troubled by this question. In his reply to the critique mentioned above, he stated as follows: “Based on what I know from the events in the 1980s as well as anecdotal evidence... I do believe that while in many ways Kosovo was a very different place than Croatia and Bosnia, the outlines of the story there were strikingly similar” (Gagnon 2007: 223). The present chapter challenges such an argument. It will argue that the dynamics of conflict in Kosovo and Macedonia were indeed quite different from other cases such as Croatia or Bosnia, because the Albanian rebellions in Kosovo and

Macedonia were directed by the “decisions from below,” and they cannot be explained by the Albanian elite’s desire to demobilise the opposition. On the contrary, as discussed below, the Albanian political elite both in Kosovo and

90 Macedonia were strongly opposing to the violence. The present thesis agrees with Gagnon on the top-down nature of the violence regarding the Serb rebellions in Croatia (and Bosnia, though Gagnon does not analyse it). This does not mean, however, that his argument is equally applicable to the cases of

Kosovo and Macedonia.

In order to achieve the aims stated above, this chapter will systematically examine each case. As for the Serbs in Croatia and Bosnia, it will examine who actually made a decision to take up arms, what kind of resources they had at their disposal, and how they used them to organise and realise the rebellion. It will finally point out that the initial level of intensity was quite high in these two cases. As for the Albanians in Kosovo and Macedonia, it will firstly examine why the Albanian elite did not choose to take up arms. It will then examine who actually made a decision to take up arms, and will show that the resources available to them were fairly limited. It will finally show that the initial level of intensity of rebellions was low in these two cases, as expected by the theoretical arguments made in Chapter 3.

1. Serbs in Croatia

(1) Involvement of Politicians and Police Forces in the Rebellion

As outlined in Chapter 2, inter-ethnic relations deteriorated rapidly after the first multi-party elections in 1990, when the Croatian Democratic

Union ( Hrvatska demokratska zajednica, HDZ hereafter), an opposition party led by Franjo Tudjman, won a landslide victory. In this context of post-election politics, it was Serb politicians from the Serb Democratic Party {Srpska demokratska stranka, SDS hereafter) who led the rebellion of Serbs against

Croatian authorities. The Serb elite from the SDS exploited the political and

91 military resources available to them. They especially utilised their political positions at the local (municipality) level to mobilise and organise Serbs. At the republic level, the SDS only got five seats in the Croatian national parliament and it withdrew its members from the Croatian parliament soon after the elections. The resources available to the SDS politicians at the republic level, therefore, were very limited. However, the SDS politicians could still use resources available at the local (municipality) level, especially where Serbs constituted a majority of the population.1

What kind of resources were available to the SDS politicians, and how did they use them to organise a rebellion? Three points can be made in this regard. Firstly, they could make a political decision in the municipal parliament

(iSkupstina opcina, SO hereafter) where they control a majority of the seats.

This enabled the SDS politicians to implement their platform as official policies of the municipality and give a democratic legitimacy to their actions.

The SDS politicians used this power to create a territorial basis of the rebellion. For example, Milan Babic, one of the leaders of the SDS who was elected as a mayor of Knin after the 1990 local elections, proposed an establishment of a new association of municipalities, and the SO of Knin authorised it.2 When the

“Serb Autonomous Region Krajina” (Srpska autonomna oblast Krajina, SAO

Krajina hereafter) was established at the end of 1990, the SO of the eight municipalities around Knin approved the Statute of the SAO Krajina.3 This

1 In the 1990 local elections in Croatia, the SDS gained a majority o f the seats in the parliament only in three municipalities. However, the SDS expanded quickly to other municipalities after the elections and gained a control o f the municipal parliament in many o f the municipalities with the Serb majority population. The results o f the local elections and the expansion o f the SDS after the elections are analysed in detail by the author elsewhere. See Kubo 2007. Barid 2005: 65. The SO o f Knin, Gradac and Donji Lapac decided to belong to it, which led to the declaration of the establishment of the “Association of Municipalities of Northern Dalmatia and Lika” on 1 July 1990. See Baric 2005:6 6 . 3 Eight municipalities that decided to belong to the SAO Krajina were Benkovac, Donji Lapac, Gracac, Knin, Obrovac, Vojnic, Titova Korenica and Dvor na Uni. See Baric 2005:

92 organisation constituted the territorial basis of the rebellion against Croatian authorities. In this regard, therefore, the political power available to Serb politicians at the municipality level played an important role in the organisation of the rebellion of Serbs in Croatia.

Secondly, the control of the SO also provided the SDS politicians with an access to arms and personnel necessary for rebellion, because the Serb politicians could control the local police force through the Secretary of Internal

Affairs (the head of the local police administration) of the municipalities.4 The local police held some weapons, and the Serb leaders could use these weapons and local policemen for a rebellion. The key figure responsible for the organisation of the Serb police force was Milan Martic, an inspector of the

Knin police. The weapons held by local police concerned the Croatian authorities,5 and they attempted to take weapons from the police station in

Serb-dominated municipalities, but this action rather provoked angry reactions from Serbs, including local policemen, who armed with the weapons of reserve police and barricaded the road and railways.6 By the end of 1990, there were several thousand Serbs armed with rifles and pistols in Knin area (Baric 2005:

83). In January 1991, the Secretariat of the Internal Affairs of SAO Krajina was established and Martic was named as a Secretary of Internal Affairs.7 At the

95. 4 This was important because Serbs constituted a large segment o f the police force in Croatia: according to the information from the Secretariat o f the Internal Affairs of the Socialist Republic o f Croatia in 1984, 49.9 % of the total employees were Serbs by nationality. See Barid 2005: 126. Note that Serbs constituted only around 10% o f the entire population in Croatia (1 1 .6 % in the 1981 census and 12.2% in the 1991 census) and thus Serbs were clearly over-represented in the police force in Croatia. 5 Alarmed by the organisation of the Serb referendum on autonomy, the Minister of the Internal Affairs o f Croatia, Josip Boljkovac, held a meeting with the secretaries of internal affairs from some municipalities near Knin. They concluded that there was a possibility that the Serb extremists could take weapons from the reserve police in the region. See Baric 2005: 78. 6 Baric 2005: 78-79. This event is often called “Log Revolution” due to the blockade of roads by logs. 7 Baric 2005: 105. The executive committee of the SAO Krajina sent a letter to

93 end of May 1991, Martic claimed that almost 20,000 military personnel were

a under his control. The organisation of police force was an “institutionalisation of the rebellion” (Baric 2005: 126). The control of the SO at the municipality level, therefore, provided the Serb leaders with some military and human resources for the rebellion.

Thirdly, the Serb politicians also used their positions to mobilise armed forces through the Territorial Defence (Teritorijalna odbrana, TO hereafter) system of the ex-Yugoslavia. The TO forces were designed to present a systematically organised guerrilla resistance to a foreign invader and were under the control of republics and municipalities, and local authorities were obliged to maintain stockpiles of weapons and supplies (Woodward 1995: 26-27). Due to this system, a significant number of small arms and light weapons were kept in each municipality.9 The Serb politicians mobilised these military resources stockpiled at each locality for their rebellion. In August 1991, the TO forces were established under the command of the government of SAO Krajina with three operative zones (Marijan 2002: 13-14). The police force organised by Martic was eventually incorporated into this defence system.

In the case of Serbs in Croatia, therefore, the Serb politicians could use the resources available to them in three ways, (1) controlling decision-making in the municipal parliaments to create the territorial basis of the rebellion, (2) mobilising the human and military resources for the rebellion

Presidency o f the SFRJ, Federal Secretary of the Internal Affairs, Ministry of Internal Affairs o f Croatia and others to notify the establishment of the autonomous secretariat for the internal affairs in Knin, to which 10 municipality police stations had entered, namely Obrovac, Benkovac, Grafiac, Titova Korenica, Donji Lapac, Dvor na Uni, Glina, Kostajnica, Vojnic and Knin. As for the text of the letter, see Letica & Nobilo 1991: 94. 8 Baric 2005: 120. Baric believes, however, that this figure is clearly exaggerated. 9 In 1989, a total of 250,000 pistols, 1,200,000 rifles and 105,000 machine guns were under die TO system. This was almost equal to the number of weapons under the control o f the regular army (the JNA), which had 270,000 pistols, 1,120,000 rifles and 170,000 machine guns. See Davis 2002: 50.

94 by controlling the local police force, and (3) controlling the TO system.

(2) Involvement of the Regular Military Force in the Rebellion

Another important aspect of the preparation of the Serb rebellion was a participation of the Yugoslav National Army (.Jugoslovenska narodna armija,

JNA hereafter). This factor also contributed to the high level of the initial intensity of rebellion. The JNA took sides with the Serbs because it was hostile to the non-Communist and “anti-Yugoslav” government in Croatia (as well as in Slovenia). It thus raided the TO stations in Croatia to seize arms stored for the Croatian TO forces immediately after the first multi-party elections in 1990

(Spegelj 2001b: 120; Silber & Little 1996: 105-118). It also started providing tacit support to Serbs in Croatia as early as in the summer of 1990: when the

Croatian authorities sent special police force to Knin by helicopters in order to de-blockade the roads, the leader of the General Headquarters of the JNA called Zagreb and warned, without knowledge of the Federal Presidency nor the Federal Prime Minister, that the army airplanes would destroy the Croatian helicopters if they proceed towards Knin (Dekanic 2004: 62; Spegelj 2001b:

127). It was also observed that the JNA provided arms to the Serb population

(Marijan 2004: 42). According to Spegelj, a number of JNA generals visited the

Serb villages in Croatia before the outbreak of the war as “emissaries and instigators of the armed insurrections in Croatia” and revived old traumas of the Second World War (Spegelj 2001a: 26).

Furthermore, the JNA actively participated in the rebellion once the

Serbs decided to take up arms. When the Serb rebels in Croatia started their military actions at the end of June 1991 onwards, the JNA openly stood on the

Serb side and started to be involved in the war.

Let us take a closer look at the size and extent of the engagement of

95 JNA forces in the war in Croatia. Firstly, the JNA mobilised the best and strongest “elite” force for the operations in Eastern Slavonia, as Veljko

Kadijevic, the then Federal Minister of Defence, explains.10 For example, in

June and July 1991, a number of elite brigades were sent from Serbia to

Vukovar area, which brought 10,000 soldiers, 450 armour vehicles and about

100 heavy weapons (Marijan 2004: 24). In July and August 1991, the forces from the 17th corps (Tuzla in Bosnia) were brought in, reinforced by the units from the 14th corps (Ljubljana in Slovenia) after its withdrawal from Slovenia

(Tus 2001: 54). The all-out offensive against Vukovar began on 24 August with air and artillery strikes (Tus 2001: 54). Secondly, as for the Western Slavonia region, the JNA forces in Bosnia (the Banja Luka Corps) were sent in and attacked OkuSani, Nova Gradiska and other places (Barid 2005: 121; Marijan

2002: 15). Thirdly, in the northern Dalmatia and Lika region, the JNA forces stationed in Knin area (9th Knin Corps led by Colonel Ratko Mladid) were deployed and attacked Zadar, Sibenik and other Dalmatian cities.11 Fourthly, the JNA navy imposed the naval blockade of the Croatian ports on 15

September 1991, even though it was relaxed after only one week due to the lack of sufficient resources (Gow 2003: 156). Finally, the JNA forces that were sent from Trebinje (Bosnia) and Herceg Novi (Montenegro) attacked the southern Dalmatian region and bombarded the old town of Dubrovnik.12

10 In explaining the plans o f the JNA operations in Croatia, Kadijevid wrote as follows: “with the strongest formation of armoured-mechanized forces (we planned to) liberate the Eastern Slavonia, and then quickly continue the action toward the west, join the forces in the Western Slavonia and go on to Zagreb and Varazdin”. See Kadijevic 1993: 135. 11 Dekanic 2004: 6 6 . As for the detailed description of the events in this region, see Kaleb 1999; Simac 2001. 12 For the discussion on the nature of the operations against Dubrovnik, see Gow 2003: 164-165.

96 (3) High Level of the Initial Intensity of Rebellion

As discussed above, the Serb politicians in Croatia could mobilise various human and military resources to prepare, organise and realise a rebellion. In addition, a large number of the Yugoslav regular army supported the Serbs and took part in the military operations against the Croatian government. With these enormous resources at their disposal, the Serb leaders in Croatia could easily make a large-scale rebellion when they decided to take up arms. While there were some skirmishes between the local Serb units and the Croatian police force after January 1991, the key moment for their decision to take up arms was the declaration of independence by the Parliament of

Croatia on 25 June 1991. After that, the government of SAO Krajina issued an ultimatum for the unconditional and immediate withdrawal of all Croatian police and military forces from the territory of SAO Krajina (Baric 2005: 120), and clashes between the Serb police force and the Croatian forces started. After

July 1991, as described above, the JNA forces also started their military operations.

Facing such a large-scale military action, the Croatian authorities were forced to react by military means, even though the Croatian political leaders did not consider the military option as the principal one at the beginning of the armed conflict (Tus 2001: 47). In July, the Croatian government issued several decrees to prepare for military operations and the General Headquarter of the

Croatian Army was finally formed on 3 August 1991 (Tus 2001: 47). The armed conflict between the Croatian forces and Serb forces including JNA units continued until January 1992 when the Sarajevo ceasefire came into force and UN peacekeeping forces arrived in March 1992. More than 6,000 people died during the war in Croatia, and another 13,700 were “missing,” the

97 majority of whom are believed to be dead.13 About 210,000 houses were destroyed and more than 300,000 people were displaced by the end of 1991

(Tanner 2001: 278; Baric 2005: 125).

2. Serbs in Bosnia

(1) Involvement of Politicians in the Rebellion In the case of the Serbs in Bosnia, it was also the politicians, from the party of the same name as in Croatia, the Serb Democratic Party of Bosnia and

Herzegovina {Srpska demokratska stranka BiH, SDS-BiH hereafter), who organised the rebellion and decided to take up arms. Compared to their counterparts in Croatia, the Serb politicians in Bosnia were even more powerful because of their electoral success and their participation in the coalition government. In the first multi-party elections in 1990, the three nationalist parties made a landslide victory, and the SDS-BiH formed a coalition government with the other two parties, namely the Party for

Democratic Action {Stranka demokratske akcije, SDA hereafter) and the Croat

Democratic Community of Bosnia and Herzegovina {Hrvatska demokratska zajednica BiH, HDZ-BiH hereafter). At the republic level, the SDS-BiH won both of the two seats allocated to Serbs in the seven-member Bosnian presidency, as well as 71 out of 240 seats in the Bosnian parliament. The electoral success of the three nationalist parties was equally striking at the local level, where they won total 4774 out of the 6299 seats (75.8%) in 110 municipalities (Amautovic 1996: 118). As for the SDS-BiH, it won the absolute majority in the SO in 28 municipalities and won the largest number of

13 Tanner 2001: 278. Bari6 gives an estimation that around 16,000 Croatian soldiers, police and civilians died during the war. See Baric 2005:124-125.

98 seats among three nationalist parties in additional 10 municipalities

(Amautovic 1996:118-120).

After the elections, the SDS-BiH, together with the coalition partners, started the process of “purging the old cadres” (Goati 1992: 58) of the former communist party in order to control the administrative organs and the resources both at the national and the local level. The division of positions and resources occurred everywhere in Bosnia. In the field of internal affairs (police), for example, it occurred as follows: in the Ministry of Internal Affairs, the SDA took 9, the HDZ-BiH took 4, and the SDS-BiH took 5 out of 18 top posts in the ministry; at the regional level, the SDS-BiH took the top positions of the regional Centre of Public Security at 3 out of 9 centres, while the other two parties took the rest (Nilsen 2004: 287-288). The SDS-BiH politicians also started to oust former communists and control the assets of the community at the local level. The success of the nationalist parties in ousting the former communists from the local government was incredible: while the leftist parties and smaller parties occupied 1364 (21.7%) out of 6299 seats in 110 municipalities (Amautovic 1996: 118), the leftist parties managed to form a local government only in two municipalities, namely Tuzla and Vare§ (Andjelic

2003: 189).

Through the electoral victory at the end of 1990, the SDS-BiH politicians gained a significant amount of resources. How did they use these resources for the organisation and preparation of the rebellion? The method used by the SDS-BiH politicians in Bosnia was strikingly similar to that of the

Serb politicians in Croatia. Given the strong connection between SDS politicians in Croatia and Bosnia, it is certain that the SDS-BiH politicians in

Bosnia followed the example of the Serbs in Croatia to mobilise and utilise their resources.

99 Firstly, the SDS-BiH politicians in Bosnia used their resources to create a territorial and institutional basis for the rebellion. After September

1991, the SDS-BiH politicians started establishing separate institutions both at the republic and the local level. At the republic level, the SDS-BiH politicians withdrew from the parliament in October 1991 as a protest against an adoption of the document “Memorandum on the Sovereignty of the Bosnia” in the

Bosnian parliament. After that, the SDS-BiH politicians established the “Serb

Parliament of Bosnia” on 24 October 1991. In December 1991, this “Serb

Parliament” elected the government (Nilsen 2004: 300). At the local level, the

Serb politicians started creating the SAOs as they did in Croatia.14 The

SDS-BiH politicians used their positions and resources available at the local level to a maximal extent: out of the 28 municipalities where the SDS-BiH took the majority of seats in the SO, all but one decided to participate in some

SAO formed in this period.15 In January 1992, the SDS-BiH leaders declared an establishment of the “Serb Republic of Bosnia” by unifying all SAOs. The political power they gained at the republic and the local level, therefore, was used to create the institutional and territorial basis for the rebellion.

Secondly, the SDS-BiH politicians mobilised the human and military resources available to them to prepare for the rebellion. They started preparations in the latter half of the year 1991, and used the local police force and the TO system as their counterparts did in Croatia. In October 1991, for example, there were some reports in the Bosnian Ministry of the Interior

Affairs that the Serbs in the Ministry secretly attempt to supply arms and help forming the SAO (Nilsen 2004: 296). In December 1991, the SDS-BiH issued

14 From September through November 1991, the SDS politicians established six SAOs, namely the SAO Hercegovina, the SAO Bosanska Krajina, the SAO Romanja, the SAO Sjeveroisto&ia Bosna, the SAO Sjevema Bosna, and the SAO Bira£. See Begic 1997: 58-59; Gow 2003: 127. 15 According to the author’s calculation, based on Amautovic 1996.

100 the “Instruction for the organisation and the activity of the organs of Serb nation in Bosnia in the extraordinary circumstances,” which envisioned the formation of the “crisis headquarters” (Nilsen 2004: 301). These crisis headquarters were formed as “para-governmental structures” (Gow 2003: 122) which made preparations for political and military aspects of the takeover. For example, the SDS-BiH headquarters in Konjic distributed around 400 weapons to local Serbs (Gow 2003: 123). The preparations made by these crisis headquarters also included ensuring the readiness of active and reserve members of the police forces, the TO and Civil Defence units. These local crisis headquarters, working closely with the military, took over the police and civil administration later (Gow 2003: 130).

(2) Involvement of the Regular Military Force in the Rebellion

In the preparations for the armed rebellion in Bosnia, the JNA also played an important role. According to Divjak, the arming of Serbs had begun in some areas as early as in 1990, and the JNA had distributed about 51,900 firearms to Serb volunteer units and 23,298 firearms to members of the

SDS-BiH by March 1991 (Divjak 2001: 154). The JNA started mobilising

Bosnian Serbs when it started participating in the war in Croatia. For example,

General Nikola Uzelac, commander of the Banja Luka corps, ordered the general mobilisation of the population of northwest Bosnia.16 According to

General Kadijevid, the mobilisation of Bosnian Serbs was quite successful and helped make up for the failure in mobilising Serbs elsewhere for the war in

Croatia (Kadijevid 1993: 147).

The JNA not only helped arm and mobilise Bosnian Serbs but also

16 The Bosnian government announced that such an order of the general mobilisation was “illegal” and President Izetbegovic called upon the Bosnian citizens not to respond to the call-up. See Burg & Shoup 1999: 83; Boji6 2001: 318-319; Tus 2001: 57.

101 was actively involved in the armed conflict in Bosnia and started fighting on the Serb side. The Serbs thus acquired an enormous amount of arms and personnel, since a large number of JNA forces were deployed to Bosnia, without an agreement of the Presidency of Bosnia Herzegovina, after the end of the war in Croatia: in April 1992, the 4th, 5th, 13th and 17th corps and the

5th air force corps were all stationed in Bosnia; the Zadar artillery training centre was moved to Sarajevo; a part of the Knin corps arrived in Bihac

(Divjak 2001: 155). General Ratko Mladic took over the Sarajevo Command in

January 1992 (Gow 2003: 77) and was appointed the commanding officer of the Army of Republika Srpska (Vojska Republike Srpske, VRS hereafter) in

May 1992 (Tus 2001: 155). According to the estimates of Gow, the VRS inherited 60,000-80,000 strong force as well as 300 tanks, 200-300 armoured vehicles and 500-600 artillery pieces.17 The VRS also inherited an air capacity of 5000 personnel and 35 aircrafts, even though the imposition of an “air exclusion zone” by the UN restricted use of this air capacity (Gow 2003: 77).

These military resources were fully used by the Serbs when the war started in

March-April 1992.

(3) High Level of the Initial Intensity of Rebellion As discussed above, the Serb leaders in Bosnia had vast political and military resources at their disposal, which enabled them to start a large-scale rebellion when they decided to take up arms. The key moment for their decision was the declaration of independence by the Bosnian government. On 3

March, based on the result of the referendum in Bosnia, Izetbegovic declared the independence of Bosnia (Burg & Shoup 1999: 118). Three days later,

17 Gow 2003: 77. Divjak gives even higher estimate of the 90,000-100,000 well armed men, cadets and officers as well as 750 to 800 tanks, about 1,000 armoured vehicles and more than 4,000 mortars and artillery weapons. See Divjak 2001: 155.

102 Karadzic called for an army takeover and warned that there would be civil war if the EC were to recognise Bosnia-Herzegovina (Burg & Shoup 1999: 118). In

March 1992, armed confrontations occurred in many places in Bosnia, such as

Bosanski Brod, Sijekovac, Doboj region and Derventa, particularly between

Croat paramilitaries and JNA reservists or Serb paramilitaries (Burg & Shoup

1999: 119). On 4 April, the JNA and Serb paramilitaries started shelling the suburbs of Sarajevo and the city itself (Divjak 2001: 157). On 6 April, when the EC granted recognition to Bosnia, the JNA forces and the Serb paramilitaries attacked Foca and began the mass killings and expulsions of

Muslim population (Divjak 2001: 157).

Facing such actions taken by the Serb side, the Bosnian government was forced to take necessary measures urgently, since the Muslims politicians from the SDA “neither believed war would break out, nor prepared their constituents in any way for its possibility” (Hoare 2004: 43). On 3-4 April, in response to the flight of Muslim refugees and under pressure from other members of the Presidency, Izetbegovid permitted the municipalities to mobilise their TO forces (Hoare 2004: 50). On 8 April, the Presidency of

Bosnia decided to form the Territorial Defence of the Republic of

Bosnia-Herzegovina, which was put into effect rapidly (Divjak 2001: 158).

Since April 1992, Bosnia was plunged into a full-brown civil war, which lasted for three years until the Dayton agreement. While the number of casualties is a matter of debate, most recent research of the Research and Documentation

Centre in Bosnia suggests that approximately 100,000 deaths were caused by the military activities, torture or detention during the war in Bosnia, of which more than 45,000 (46%) occurred in the initial year (1992) alone.18 The

18 The details of the research conducted by the Research and Documentation Centre can be obtained from their website at: http ://www. idc. org.ba/presentation/content.htm.

103 estimates of the total number of displaced persons range from one-third to two-thirds of the total pre-war population in Bosnia, or from over 1.4 million to

2.5 million people (Burg & Shoup 1999: 171-172).

3. Albanians in Kosovo

(1) Path Chosen by the Albanian Elite in Kosovo: Peaceful Resistance

Albanians in Kosovo started mobilising themselves against the

Serbian government in the late 1980s, primarily as a response to the repressive policies implemented by Milosevic. Before the rise of Milosevic in 1987,

Serbian nationalism was suppressed by the Serbian Communist authorities led by Ivan Stambolic.19 The situation, however, was drastically changed by

Milosevic’s visit to Kosovo in 1987, when he turned into a nationalist.20 After his victory in the internal conflict within the Serbian communist party,

MiloSevid began introducing a series of repressive policies. The exclusion of

Albanians from the political and economic life was conducted “legally,” based on the amendments of the Serbian constitution in 1989, a number of laws enacted by the Serbian parliament after 1989, and the new constitution of the

Republic of Serbia introduced in 1990 21 While Albanians made a number of protests against these policies,22 the Serbian authorities took a tough stance

19 For example, when the “memorandum” which carried Serbian nationalist claims was reported in the newspaper “VeCemje Novosti” in 1986, the Serbian authorities severely criticised these nationalist claims. The media in Serbia also criticised them so severely that the media in Slovenia praised the anti-nationalist stance taken by the Serbian media. See Silber & Little 1996: 33. 20 When thousands of angry Serbs were clashing with the Albanian police and throwing stones at them, MiloSevid came out o f the building and said “No one should dare to beat you!” The clip o f Milosevic saying this phrase was shown by the Beograde Television over and over again. 21 For the analysis of these legal documents and the discrimination based on these laws, see HCHRS 1997; HCHRS 1998. 22 For example, in 1988, when Albanian leaders were forced to resign, two thousand Albanian miners from Trepca marched to Prishtina, where tens o f thousands of students

104 against these demonstrations. In March 1989, for example, Serbian authorities issued the state of emergency and mobilised security forces to repress demonstrations, leading to the casualties of 24 lives. In January 1990, clashes between demonstrators and the Serbian security forces culminated in the death of 28 people and the injuries of 94 people.

Facing such a tough stance of the Serbian authorities, Albanian elite started making more secessionist demands. In July 1990, for example, the

Albanian MPs of the Parliament of Kosovo unilaterally declared that Kosovo become a republic and independent from Serbia, even though it still remains a part of Yugoslav federation.23 At the end of September 1991, they organised a referendum on “sovereignty and independence of Kosovo,” where the voter turnout was 87% and 99.87% of those who voted supported it. In May 1992,

Albanians organised their own presidential and parliamentary elections, and the

Democratic League of Kosovo (Lidhja demokratike e Kosoves, LDK hereafter) won the overwhelming majority (76.44%) and its leader, Ibrahim Rugova, became president of Kosovo. Led by these elite, Albanians in Kosovo began building a “parallel society” (e.g. Kostovidova 1996). For example, Albanians established a “parallel” system of education in the Albanian language. In order to sustain their parallel system, the Albanian elite established their own “tax” system, in which not only local Albanians but also diaspora communities in the

Western Europe paid some money to finance these activities. Albanians were thus very well mobilised and organised under the leadership of the LDK by the end of 1992. and others joined them, to protest in front of the party headquarters. In February 1989, more than one thousand Trepca miners began a hunger strike, and demonstrations spread across Kosovo. When constitutional amendments were approved by the Kosovo parliament controlled by Milosevic in March 1989, huge demonstrations occurred throughout the province. 23 For the English translation of the text of “Constitutional Declaration of the Assembly of Kosova (Prishtina, July 2,1990)”, see ASRA1993: 331.

105 This situation, however, did not lead to a “decision from above” to take up arms in Kosovo, because the LDK adopted non-violent strategies and did not choose to organise an armed rebellion. Even when Croats and later

Bosnian Muslims encouraged Albanians to take up arms and open the

“southern front” against Serbia, Albanian leaders rejected their request (Judah

2000a: 113-115). Why did they choose peaceful protest rather than violent means? There are several factors that affected their decision. Firstly, there was an influence of the ideas of democratic opposition in Eastern Europe: the idea of a parallel system or a “shadow” government was influenced by the notions of autonomy and self-organisation developed among Central European intellectuals, particularly Polish Solidarity (IICK 2000: 44-45). Secondly, there was a move among ordinary Albanians to counter the “primitive and uncivilized” stereotype portrayed by the Serbs: Albanians decided to abolish the traditional practice of blood feud, establishing a “Council of

Reconciliation” which tracked down Albanian families and brought them together for mass reconciliation (IICK 2000: 44-45). Thirdly, there was a pragmatic judgment as well: the Serbian armed forces were too strong to fight against. In 1992, for example, Rugova made the following remark: “we would have no chance of successfully resisting the army... in fact the Serbs only wait for a pretext to attack the Albanian population and wipe it out. We believe it is better to do nothing and stay alive than be massacred” (Vickers 1998: 264).

Fourthly, the attitude of external actors was also important. Albanian president

Sali Berisha did not support an openly violent course because he was not willing to jeopardise the economic assistance from abroad and to jeopardise the security of Albania by inviting Serbian reprisals (Judah 2000a: 115). He also openly opposed the idea of unification of Albania and Kosovo: for example,

Berisha stated in 1993 that “the idea of a Greater Albania is not considered in

106 serious Albanian political circles” (Vickers 1998: 270). Due to these factors, the non-violent movements led by the LDK had been predominant among

Albanians up to 1997 and “as late as the autumn of 1997, few foresaw the threat of an imminent and major guerrilla-led uprising in Kosovo” (Judah

2000a: 119).

(2) Decision from Below: the KLA and Its Limited Resources

While the Albanian elite in Kosovo had chosen a non-violent strategy, a small resistance group named the Kosovo Liberation Army (Ushtria qlirimtare e Kosov es, KLA hereafter) emerged and started an armed rebellion against the Serbian authorities in Kosovo. The KLA was founded in 1993 and grew out of the Popular Movement for Kosovo (Levizja popullore per Kosoves,

LPK hereafter), a Marxist underground party formed in the early 1980s.24

Those who organised the KLA were mainly the activists or students who went abroad after being punished in Serbia. For example, Hashim Tha$i, who later became a head of the KLA political directorate during the war and the Prime

Minister of Kosovo provisional government after the war, was a head of the students’ union in 1991, lived in Switzerland studying politics and international relations in Zurich in 1994-5, and worked on meetings with the Albanian diaspora (Judah 2000a: 117). Bardhyl Mahmuti, who later became a francophone spokesman of the KLA in Switzerland and a foreign minister of the Kosovo provisional government after the war, was imprisoned from 1981 to 1988 after 1981 demonstrations and lived in Switzerland from 1990 and studied political science at Lausanne (ICG 1999: 15). Pleurat Sejdiu, an orthopedist and a member of the LPK in Llap, was punished for taking part in

24 For the LPK and other Maxist-Enverist underground parties, see Judah 2000a: 102-120; £eku 2003; £eku 2004.

107 the demonstrations in 1981 and left for Bucharest to be a student there, and then was working as a mini-cab driver in London, where he emerged as a

KLA’s spokesman (Judah 2000a: 117-118; ICG 1999: 16). Xhavit Haliti, who was one of the four men who began building a network of secret cells for the rebellion, also left Kosovo in the 1980s and had been in Switzerland before he started working as a logistics and finance man for the KLA (ICG 1999: 14).

Jashar Salihu, who worked as a chairman of the Homeland Calling fund of the

KLA, was imprisoned for four years for distributing leaflets in 1981 and moved to Switzerland after that (Judah 2000a: 104-107). In a nutshell, the KLA was organised and led by radical activists, former demonstrators and students who went abroad in the 1980s or later. In Kosovo, therefore, the rebellion was led not by the political elite, such as the LDK and Rugova, but by the members of small underground organisation, and thus came from the “decision from below.”

Because the organisers of rebellion were members of a small underground organisation, the resources at their disposal were very limited when they started their rebellion activities. For example, up to 1997, the KLA had only about 150 active men, according to Mahmuti (Judah 2000a: 118):

According to the Independent International Commission on Kosovo (IICK hereafter), “until late 1997, active armed resistance groups in Kosovo were very small and without permanent bases in the province” (IICK 2000: 52). The military hardware which the KLA had at its disposal was limited to small weapons that “a mule can carry,” since all its weaponry had to enter the country illegally, mainly from Albania but also sometimes via Montenegro or

Macedonia (ICG 1998c: 6). Because the LDK had already organised the

Albanian diaspora communities in the Western countries for financial support

(according to the then Prime Minister of the govemment-in-exile of the

108 Republic of Kosova, Bujar Bukoshi, their fund had mobilised a total of 265 million DM by the end of the war in Kosovo25), the KLA members could not get much financial support from the diaspora communities, and thus financial resources were also fairly limited. According to Ibrahim Kelmendi, the LPK members contacted Bukoshi in 1992, but Bukoshi and the LDK people rejected to support the LPK members financially from their fund26 While Rugova and the LDK people thus controlled a large amount of financial resources, they did not use them for violent activities, and the KLA did not have significant financial resources at their disposal during their preparation of the rebellion.

(3) Low Level of the Initial Intensity of Rebellion As a result of the limited resources available to the organisers of rebellion, the initial intensity of rebellion was not very high in Kosovo. The first violent action allegedly taken by the KLA was the killing of a Serbian policeman in 1995, but it was not until 1996 that an organisation calling itself the KLA claimed responsibility for the attacks (IICK 2000: 51). The first

“planned” assaults took place on 22 April 1996, when four almost simultaneous attacks were launched in separate locations that killed two policemen (ICG

1998c: 2). Since then, there were some sporadic KLA attacks on Serbian policemen but the intensity of rebellion remained quite low. It was on 15

October 1997 that the first KLA man “in uniform” died while attacking a police station at KliSina (Judah 2000a: 117). On 28 November, some KLA members appeared in public for the first time at the funeral of Halit Gecaj, an

25 Interview with Bujar Bukoshi, Prishtina, 2006/05/05. 26 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30. Kelmendi went into exile in Germany in 1976 and was one of the core founders of the LPK. He is currently the political advisor of Ali Ahmeti, former leader of the NLA and currently the president of the Albanian political party in Macedonia.

109 Albanian killed in the crossfire of the KLA attack on Serbian police.27 While the number of attacks was increasing (31 in 1996, 55 in 1997, and 66 in

January and February 1998 alone), the death toll remained relatively low in this period: according to Ljubiska Cvetic, who was then a spokesman for the

Ministry of Internal affairs (Ministarstvo unutrasnjih poslova, MUP hereafter) of Serbia, these attacks from 1996 to February 1998 led to 10 deaths of Serbian policemen and 24 deaths of civilians28 One can conclude, therefore, that the

KLA started its activities from quite a low level of intensity and this low-intensity activity continued until the beginning of 1998.

Since 1998, the conflict escalated into a full-brown fight between the

KLA and the Serbian security forces, which led to the intervention by NATO in

1999 and eventually led to the large number of deaths: while the exact number of killings during the conflict is unknown, it is estimated that the number of killings was around 10,000, with the vast majority of the victims being Kosovo

Albanians killed by the Yugoslav forces and furthermore, some 3,000 are missing, many of which are estimated to be dead (IICK 2000: 91). Despite the eventual serious casualties and the high intensity of the conflict, the KLA could start its rebellion only from a quite low level of intensity, due to the limited resources available to them.

4. Albanians in Macedonia

(1) Path Chosen by the Albanian Elite: Moderation and Participation

Like their fellows in Kosovo, Albanians in Macedonia had been

27 ICG 1998a: 29. Now, 28 November is a date observed by Albanians everywhere as Flag Day, a holiday of great patriotic significance. See ICG 1998c: 3. 28 ICG 1998a: 30. As for the list of the date and place of the attacks associated with the KLA during this period, both on the Serbs and Albanian “collaborators,” see Kresovic et al 1998:40-49.

110 experiencing various kinds of repressions, though to a different extent. During the socialist era, Albanian communities in western Macedonia faced harsh political and cultural repression under Rankovic and also after the death of Tito.

After the Albanian demonstrations occurred in Kosovo in 1981, the

Macedonian Assembly took measures to crack down on Albanian nationalism. 29 Albanian civil servants were sacked from the state administration in 1987, and some ethnic Albanian army officers were disciplined for participating in Albanian weddings where Albanian nationalist songs were chanted (Phillips 2004: 45). Once the Albanians in Kosovo achieved a considerable degree of autonomy under the decentralized system of the 1974 constitution, the chief objective of the Albanian nationalist activists was to improve the status of the Albanians in Macedonia by securing the recognition of their collective rights (Qosja 1990).

While the situation improved on certain issues in the 1990s, such as secondary education,30 Albanians in Macedonia remained dissatisfied on several issues after the transition to the multi-party political system. The first issue was the census results: Albanians claimed that the results of the 1991 census underestimated the number of Albanians because many Albanians boycotted the census. This issue was resolved to some extent by the census under the observation of the Council of Europe in 1994, but many observers believe that the number of Albanians is still underestimated due to many not being recognised under the harsh citizenship requirements (Poulton 2000: 186).

29 For example, school curricula and textbooks were revised and Albanian-language schools were ordered to teach more Macedonian. See Phillips 2004: 45; Roudometof 2002: 167. 30 See Poulton 2000: 185. The number of regular secondary schools with instruction in the Albanian language increased from6 in 1991/92 school year to 19 in 95/96 year, and the number o f Albanian pupils increased from 2875 in 91/92 year (4.1% o f the total) to 8687 (11.2%) in 95/96 year. For the details of the statistics, see Milosavlevski & Tomovski 1997: 329-345.

Ill The second issue was the higher education (university level) in the

Albanian language. In the socialist era, many Albanian students attended

Prishtina University in Kosovo, which was almost entirely Albanian in intake

(Poulton 2000: 132). Since the access to Prishtina University was lost due to the suppressions by the Serbian authorities, Albanians in Macedonia demanded an official approval of the university in the Albanian language and unilaterally set up “Albanian University” in Tetovo, which started to function illegally in

1995 (Poulton 2000: 185). The Macedonian authorities, however, kept refusing demands for the official approval of Tetovo University.

Thirdly, the display of the Albanian national symbols was also a contentious issue. When newly-elected mayors of some municipalities raised the Albanian flag over the municipal offices, the Macedonian authorities ruled that it was illegal and the police moved in to enforce the removal of the flag, which led to three deaths of protesters and the arrest of the mayor of Gostivar and 400 others (Poulton 2000:189).

Finally, the lack of equal opportunity for employment, especially in the public sector, was also an important issue. Albanians claimed that they were underrepresented in the public sector: in the police and the army, for example, Macedonians constituted 93.9% whereas Albanians constituted only

3.1% (ICG 2000: 17-18). In 1993, a leader of the Albanian party in Macedonia pointed out that there were no courts presided over by Albanians, no ethnic

Albanians in the Macedonian Army General Staff or in the Interior and Foreign

Ministries (Poulton 2000: 188). Overall, Albanians claimed that they were reduced to second-class citizens and they demanded to be recognised as a

“constituent nation” of Macedonia.

These grievances among Albanians, however, did not lead to the

“decision from above” to take up arms in Macedonia, because the Albanian

112 politicians did not choose to rebel. For example, when some local Albanian politicians announced the establishment of a self-proclaimed Albanian autonomous republic in April 1992, the leaders of the major Albanian party severely criticised this move (Bugajski 1994: 116-117). When the Kosovo crisis broke out in 1998, an influx of Albanian refugees into Macedonia occurred and the Macedonians were on alert, but the Albanian elite did not take any opportunistic action to destabilise the country.31 This moderate attitude continued even after the revolts of the Albanian rebels in Macedonia in 2001: leaders of the Albanian parties criticised the armed rebellion and called for the rebels to put down arms.32

This relatively moderate attitude of the Albanian politicians was partly due to the willingness of the Macedonian parties to form a power-sharing coalition government with Albanian parties: since 1990, Macedonian parties that won the elections always offered some Albanian party to be a member of the coalition government and to take some ministerial posts. Such attitudes of the Macedonian parties certainly contributed to the moderation of the

Albanian politicians in Macedonia. In addition, international factors also played an important role. Since the Albanians in neighbouring Kosovo had chosen a non-violent path, they did not encourage the Albanians in Macedonia to take up arms. As discussed above, Albania also did not encourage fellow

Albanians in the former Yugoslavia to take up arms.

31 RFE/RL Balkan Report, 3-30,1999/08/03. 32 IWPR'S Balkan Crisis Report, 224, 2001/03/07; IWPR'S Balkan Crisis Report, 227, 2001/03/16. 33 The author examined elsewhere the continuation of the power-sharing coalition government in Macedonia since 1990, comparing it with the establishment of the exclusive one-party government in Croatia in 1990. See Kubo 2004.

113 (2) Decision from Below: The NLA and Its Limited Resources

While the Albanian political elite thus did not choose to take up arms, a small guerrilla group named the National Liberation Army (Ushtria glirimtare kombetare, NLA hereafter) emerged and started its rebellion in 2001.

The leader of the NLA, Ali Ahmeti, was one of four men who prepared to set up the KLA in Kosovo, known by his pseudonym “Abaz Xhuka” at that time.34

Ahmeti was a member of the radical Marxist-Leninist group and involved in the demonstrations in Kosovo in 1981, and sentenced to six months imprisonment, after which he fled to Switzerland (Phillips 2004: 8). Another leader of the NLA who emerged at the early phase of the rebellion, Fazli Veliu, was his uncle and a former high-school teacher (Phillips 2004: 8). Veliu was wanted by the authorities in Skopje in connection with terrorist bombing attack in his home town of Krcevo in 1998, while he denied the charge.35 The NLA was thus organised by a small number of members of an underground organisation. In other words, the rebellion by the NLA in Macedonia was directed by the “decision from below,” rather than the “decision from above.”

Just like their counterparts in Kosovo, the organisers of the rebellion had fairly limited resources at their disposal. By May 2001, for example, it was estimated that the NLA had 800 men under arms (Phillips 2004: 111). The NLA commander in one region stated that his units had no mortars above 82mm

(Phillips 2004: 106).

An important factor that limited the resources available to the NLA was the lack of systematic support from the Albanians in Kosovo. Certainly,

Ahmeti and Veliu were heavily involved in the KLA activities. This does not mean, however, that the KLA or the LPK people systematically decided to

34 Philips 2004: 8 . As for the four-man branch of the LPK, see also Judah 2000a: 115-116. 35 IWPR'S Balkan Crisis Report, 227,2001/3/16.

114 organise another “liberation struggle” in Macedonia. On the contrary, the LPK held a convention in Prizren in August 1999 and decided to dissolve itself now that its mission had been completed.36 For the Albanians outside Kosovo, this meant that they needed to organise their own movement again.37 While

Ahmeti later decided to establish the NLA with his friends from Macedonia and started gathering arms that remained in Kosovo, “even the best friends from the KLA, such as Thagi and Mahmuti, were against such action and they argued that it was wrong to make another rebellion.”38 Due to the lack of systematic support from the Albanians in Kosovo, the organisers of the rebellion in Macedonia could not mobilise the resources as much as they could have done if they had enjoyed full support from the Albanians in Kosovo.

(3) Low Level of the Initial Intensity of Rebellion Since the resources available to the rebels were fairly limited, the NLA could start its rebel activities only at the low level of intensity. The first NLA actions occurred in January 2001, when a police station in the predominantly

Albanian village of Tearce was attacked by grenade and one policeman was killed (Rusi 2004: 2). In February, there were gun-battles between government forces and the rebels near the village called Tanusevci (Churcher 2002: 17). In

March 2002, battles between rebels and government forces occurred around

Tetovo (Phillips 2004: 85-98). The armed conflict between NLA and the

Macedonian forces continued since the end of February until August 2001 when the “Framework Agreement” for constitutional changes (so-called “Ohrid

Agreement”) was signed by two Macedonian and two Albanian parties. While fighting did not stop immediately, armed insurrection by the NLA largely

36 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30. 37 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30. 38 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30.

115 ceased after the agreement. During six months of conflict, the level of intensity of rebellion remained relatively low, never reaching to the level of protracted civil war. In six months of fighting between the Macedonian security forces and ethnic Albanian guerrillas, it is estimated that between 150 and 250 people were killed (Phillips 2004: 161). The armed conflict in Macedonia was far short of the level of severity of other conflicts in the ex-Yugoslav region, at least in terms of the death casualties. Indeed, if one applies the common operationalisation of “civil war” in the large-N studies which takes 1,000 deaths as the threshold of the “onset” of civil war, the Macedonian conflict would not be classified as a “civil war.”

Conclusion

This chapter has examined the relationship between the type of decision-maker and the initial intensity of rebellion in four ex-Yugoslav countries. The Serb rebellions in Croatia and Bosnia were directed by

“decisions from above” and the initial intensity of rebellion was very high. On the contrary, the Albanian rebellions in Serbia and Macedonia were caused by

“decisions from below” and the initial intensity of rebellion was fairly low. The amount of resources available to the organisers of rebellion was the key factor that explains the causal mechanism: the Serb organisers of rebellion had a large amount of resources at their disposal, while the Albanian organisers only had a limited amount of resources for rebellion.

As was pointed out in Chapter 3, this argument does not mean that the type of decision-maker is the only factor that determines the initial intensity of rebellion. Another important factor is the inter-ethnic factor. It was argued in

Chapter 3 that the initial level of intensity would be higher if the ethnic group

116 is powerful, while it would be lower if the ethnic group is less powerful within the state. This seems to be an important factor that explains the difference between the Serb rebellions and the Albanian ones: Serbs were much more powerful in the former Yugoslavia than Albanians, occupying a large share of the top cadres in the JNA and having the largest Republic (Republic of Serbia) under their control. One can surely argue that this inter-ethnic difference also explains different outcomes in the initial intensity of rebellion among the cases of ex-Yugoslav countries.

The present chapter cannot examine which factor was more important.

In any case, the large-N analysis in Chapter 3 has shown that the intra-ethnic factor remains significant even when one controls for the inter-ethnic factor and other conditions. This conclusion seems to be valid for the cases of ex-Yugoslavia as well. For example, while it is true that Albanians in general had much less resources at their disposal compared to Serbs, it is also true that the Albanian political elite in Kosovo and Macedonia had much more financial and human resources than those who actually organised rebellions. Therefore, one can make a counterfactual argument39 that the initial level of intensity of rebellion would have been higher than it actually was if the Albanian political elite had decided to go to rebellion. The intra-ethnic factor thus constitutes an important part of the explanation of different outcomes in the initial intensity of rebellion.

The present thesis now proceeds to the next step of the three-step approach, namely the analysis of the causes and determinants of the initial occurrence of rebellion. This will be the task of Part III.

■5Q For the method of counterfactual argument, see Fearon 1991.

117 Part El: Onset of Initial Rebellion

Chapter 5 Etiology of Ethnic Rebellion: Quantitative Analysis of the Effect of Dynamic Grievances

Introduction

This chapter addresses a fundamental question regarding the occurrence of ethnic rebellions - why do people decide to take up arms? In order to do so, it will firstly present a model for the explanation of the decisions by some members of the ethnic group to take up arms. This model takes into consideration both structural conditions and the dynamic element of grievances discussed in Chapter 1. The present chapter then conducts a large-N analysis based on the MAR dataset. A research design is firstly discussed and then the results will be presented and discussed.

1. Theoretical Model: Dynamic Grievances and Structural Conditions

A model presented here is based on some basic assumptions. Firstly, this model considers violence (rebellion) as an instrument to change the status quo. Secondly, this model assumes that there are other options to take for the potential organisers of the rebellion, and that they will assess other options they could take before they actually start preparing and organising rebellion. Given the organisational nature of the violence, these assumptions seem reasonable.

Normally, rebellion is not a spontaneous and irrational reaction: in order to make a rebellion into reality, one has to spend significant resources in terms of

118 preparation and organisation. Therefore, it will be realistic to assume that the potential organisers of the rebellion will assess other options they can take before they actually start preparing and organising a rebellion.

As for the other options they can take, one can think of two other fundamental options. The first is peaceful means to change the status quo.

When one attempts to change the status quo, whether one should take peaceful means or violent means is a critical choice. Peaceful means to change the status quo therefore constitute an important alternative to violent means such as rebellions. The second is to accept the continuation of the status quo and not to take any action to change it. When one decides to take any action whether it is violent or peaceful, there must be a judgment that taking an action in order to change the status quo is better than simply accepting the status quo. The model thus assumes that there are basically three options for the potential organisers of collective action, namely (1) to accept the status quo and to do nothing, (2) to take peaceful means to change the status quo, and (3) to take violent means to change the status quo.

How do they compare and assess these three options they could take?

One can make a model with some simple mathematical expressions to simplify the assessment process. Firstly, let us express the perceived value o f the continuation o f status quo in the eyes of potential organisers of collective action by SQ. Secondly, let us express the perceived and expected value of the change of status quo by SQ2. This SQ2 thus stands for the improvement from the status quo, such as the elimination of discriminations or the restoration of autonomy. By definition, therefore, one can express the relationship between

SQ and SQ2 mathematically as SQ2 > SQ. Any collective action to change the status quo could occur only when such SQ2 is conceivable. In other words, if the status quo is so good for the members of the ethnic group that one cannot

119 conceive any SQ2 which is higher than SQ, no collective action will be made by the members of the ethnic group.

When one decides to take an action to change the status quo, one considers the cost of action. Let us thus express the perceived and expected cost of the peaceful protest option by CPROT and the perceived and expected cost of the violent rebellion option by CREB. As long as the cost of at least one type of action is smaller than the difference in the value between SQ and SQ2, one would take a less costly option to achieve SQ2. On the other hand, if the cost of both types of action is larger than the difference in the value between

SQ and SQ2, one would not take an action even when such SQ2 is indeed conceivable. In this case, people do not take any action not because they are fully satisfied with the status quo but because they think that taking an action to change the status quo is too costly and that it is better off not to take any action even though such SQ2 is theoretically conceivable.

As long as SQ2 is conceivable, therefore, three different decisions made by the potential organisers of the rebellion can be predicted according to the relative values of four factors in the following manner. One would make:

(1) No action if SQ > (SQ2 - CPROT) and SQ > (SQ2 - CREB)

(2) Peaceful protest if CPROT < CREB and (SQ2 - CPROT) > SQ

(3) Violent rebellion if CREB < CPROT and (SQ2 - CREB) > SQ

An equation (1) stands for the situation where both types of actions are perceived as too costly given the improvement one can expect by making such an action. This situation could occur either because the cost is perceived to be very high or because the improvement one can make (SQ2 - SQ) is not so large. An equation (2) stands for the situation where it is perceived that the peaceful protest is less costly than the violent rebellion and that the cost of peaceful protest is not too high given the improvement one can expect by

120 making such an action. An equation (3) stands for the situation where it is perceived that the violent rebellion is less costly than the peaceful protest and that the cost of violent rebellion is not too high given the improvement one can expect by making such an action. Based on this simple model, the following part will try to develop theoretical arguments and draw hypotheses to explain the onset of the rebellion.

(1) Structural Conditions

Let us firstly examine a set of structural conditions that would determine the onset of rebellion. Chapter 1 argued that an overwhelming majority of the existing literature on the causes of the occurrence of rebellion remain static, focusing on the structural conditions. The model presented above does not deny the importance of structural conditions. On the contrary, one can incorporate a variety of structural conditions into the explanatory framework.

The model discussed above implies two general principles related to the structural conditions. Firstly, if the structural conditions lower the perceived value of the status quo in the eyes of members of the ethnic group, they will increase the probability that some members of the ethnic group will decide to go to rebellion, other conditions being equal (the lower the value of SQ, the higher the probability of occurrence of ethnic rebellion). Secondly, if the structural conditions lower the perceived cost of rebellion, they will increase the probability that some members of the ethnic group will decide to rebel, other conditions being equal (the lower the value of CREB, the higher the probability of occurrence of ethnic rebellion). One can call the former

“motivating conditions” and the latter “enabling conditions.” In the present analysis, these two types of structural conditions are considered, mostly based on the key findings in the existing literature.

121 Contrasting the motivations and enabling conditions as causes of conflict is not new. Indeed, a large class of theories of ethnic conflict explains war as a product of motivating and enabling conditions (Oberg 2002: 15).

However, some scholars tend to see these two types of conditions as mutually exclusive and argue that only one type is important for the occurrence of internal conflict. On the contrary, the model presented in this thesis suggests the complementary nature of these two types and the possibility of different combinations of factors for the explanation of ethnic rebellion. The model presented above suggests that, in some cases, ethnic rebellion may occur even when there are no enabling conditions if the status quo is so bad. The model also suggests that, in other cases, ethnic rebellion may occur even when the status quo is not so bad if opportunities for rebellion are so manifest. According to this model, what is important is the relationship between the status quo and the opportunities for change.

Motivating Conditions Motivating conditions relate to the motives of the potential rebels.

Clearly, there would be no rebellion if there are no conditions that motivate members of ethnic groups to take up arms. In the model presented above, this logic could be expressed as follows: the higher the SQ, the more likely that the members of ethnic groups are satisfied with the SQ; and the lower the SQ, the more likely that the members of ethnic groups will find a change of the status quo (SQ2) more desirable even when the cost of the action is taken into account. In other words, conditions that lower the value of SQ can be regarded as motivating conditions.

What conditions are likely to lower the perceived value of SQ for the members of ethnic groups? There are at least two such factors. The first factor

122 is the presence of discrimination against the ethnic group. In the literature on ethnic conflict, discrimination is often cited as a source of grievances that motivate members of the ethnic group to take up arms. Analysing ethnic secessionist conflict, for example, many authors have argued that discrimination against ethnic groups provide a strong incentive. Sterling argues that the motive of separatism is almost invariably the conviction that the population involved is the victim of discrimination (Sterling 1979). Heraclides

(1991: 17), who examined various secessionist conflicts in the world, argues that “[f]or separatism to be bom there should exist some kind of disadvantage.”

As discussed in Chapter 1, Gurr (2000: 105) also argues that, for disadvantaged groups, “invidious treatment and repression are primary incentives for ethnopolitical action.” Brown (1997) argues that discriminatory political institutions and discriminatory economic systems are often identified as one of the underlying causes of internal conflict. One can expect, therefore, that the presence of discrimination against the ethnic group would lower the value of SQ for this ethnic group and thus increase the probability of the occurrence of rebellion.

The second factor is the loss of autonomy. The statehood or autonomy is often of primary importance both practically and symbolically for the ethnic group, and the loss of the statehood or autonomy can cause widely-held grievances among members of the ethnic group. Coakley (1994: 1) points out that the prominence of territorial demands in the rhetoric of ethnic activists is an extremely common phenomenon, such as demands for autonomy within a state, for separation from it, or for unification with another state. Conflicts in which groups claim exclusive title of territory for control or access are generally perceived to be more intractable, since conflict over territory is usually zero-sum, especially when territory is associated with symbolic or

123 cultural aspects of group solidarity (Sisk 1996: 18). Bose (2002, 2003) also argues that the most intractable ethnic conflicts are often caused by the conflicting claims of self-determination held by the ethnic groups. Therefore, the loss of independence or autonomy, caused by the forced annexation or other reasons, can lower the value of SQ and motivate members of the ethnic group to take up arms.

Enabling Conditions

Enabling conditions are related to what have often been termed

“opportunities.” The fundamental reasoning here is that the rebellion would be impossible, even if the potential rebels are motivated, if there are no conditions that enable the rebellion. In the model presented above, this logic can be expressed as follows: even if the value of SQ is low, making a rebellion to change the status quo would not be a desirable option for members of the ethnic group if it is so costly. What conditions are, then, likely to affect

(increase or decrease) the perceived cost of rebellion? Five such factors are considered, namely (1) type of the political regime, (2) strength of the government authority, (3) the level of economic development, (4) geographic terrain of the state, and (5) regional concentration of the ethnic group.

The first factor is the nature of the political regime. If the political regime is a harsh authoritarian one, one would expect a harsher response by the state authorities to violent collective action. If a state is under such a political regime, furthermore, associational and organisational activities of the people are often severely restricted by the political regime and thus it would be more difficult for the potential rebels to organise rebellion activities. One could expect, therefore, that the cost of rebellion would be higher under the harsh authoritarian regime.

124 The second factor is the weakening of government authority. This may decrease the cost of rebellion (perceived by the potential rebels) because one would expect weaker and less harsh responses by the state authorities if government authority is weak or weakening. As discussed in Chapter 3, this argument has been supported by statistical analysis of the onset of civil war conducted by Fearon and Laitin (2003). Following them, two indicators of the weakening of the government authority will be considered in the present analysis again, namely the political transition and the newness of the state.

The third factor is the level of economic development. As discussed in

Chapter 3, it has been shown by a number of studies that the level of economic development has a significant effect on the occurrence of civil wars. Collier and Hoeffler believe that this factor represents the “opportunity cost” of rebellion for individuals. Fearon and Laitin argue that this factor represents the weakness of the state: the lower the economic development is, the smaller the economic resources that the state can use against the rebels. In either way, higher level of economic development is regarded as a higher cost for the rebellion. One would thus expect that it would be less likely that members of the ethnic group in the country decide to take up arms as the level of the economic development in the country becomes higher.

The fourth factor is the geographic conditions of the state. Fearon and

Laitin have emphasised the importance of the “technology of insurgency” and shown that the mountainous terrain is related to the occurrence of civil war: the mountainous terrain helps the rebels to hide from the government forces and makes it easier for them to conduct guerrilla-insurgency warfare against the state. One would expect, therefore, that the perceived cost of rebellion is lower for the members of the ethnic group if a country is more mountainous.

The fifth factor is the regional concentration of the ethnic group in the

125 state. If the ethnic group is concentrated in the region, this means that the potential rebels could have a regional and territorial basis for the rebellion. On the other hand, if the group is primarily urban or dispersed throughout the country, the potential rebels cannot have such a regional and territorial basis for the rebellion. If they do not have any regional and territorial basis for the rebellion, it would be difficult for them to fight against the government forces effectively or to protect them from the attacks of the government forces. One could thus expect that the perceived cost of rebellion is lower for the members of the ethnic group that is regionally more concentrated within the state. Toft was the first to stress the association between the group concentration and the rebellion in the MAR dataset (Fearon & Laitin 1999: 16), and Fearon and

Laitin (1999: 16) also found empirically that this factor is “powerful and robust” and “minorities that have a rural base are far more likely to see large-scale ethnic violence than urban and widely dispersed minorities.”

Therefore, one would expect that it is more likely that members of the ethnic group decide to take up arms if the ethnic group is regionally concentrated.

(2) Dynamic Grievances: Importance of the Time Dimension

The preceding section discussed a set of structural conditions that may affect the occurrence of ethnic rebellion, based on the key findings and arguments in the literature. Now, let us assume for the moment that the values of SQ, SQ2, CPROT and CREB will be determined by these structural conditions discussed above. If one does not take into account the time dimension, the predicted outcome will be the same as far as the conditions that determine the value of these factors remain unchanged. In other words, if one chooses to take one option at some time, t=0, the model without time dimension will predict that one will also choose to take the same option at the

126 next period, t=l, as long as the conditions that affect the value of these factors remain the same.

This prediction is, however, not plausible. Let us examine a hypothetical situation where some members of the ethnic group decide to go to peaceful protest in order to change the status quo at some time, say t=0. For the discussion here, it is assumed that the ethnic group cannot change the status quo alone and the state authorities must decide whether they will change the status quo according to the demands of the ethnic group. If the state authorities do change the status quo, this move creates a new status quo at the next moment, t=T. However, if the state authorities ignore the ethnic group’s demands and decide to maintain the status quo, the status quo remains at the next moment, t=l, and the organisers of collective action will have to consider what to do again under the exactly same conditions as the previous moment. As discussed above, if one does not consider the time dimension, the prediction will be the same for the next moment, t=l, since all factors remain the same: they will make a peaceful protest at t=l again. If the state authorities keep ignoring their demands, therefore, there will be an endless repetition of peaceful protests by the ethnic group and their rejection by the state authorities.

This endless repetition of peaceful protests and their rejection, however, is not a realistic prediction because members of the ethnic group would gradually learn that peaceful protests are not effective. As time goes by, members of the ethnic group will get frustrated by the fact that peaceful protests failed to achieve a change of the status quo, and will look for other options to achieve their goal. In order to incorporate this dynamism into the model, let us introduce the discounting factor p and assume that the expected and perceived value of the SQ2 achieved by peaceful protests will be discounted by this factor p (where 0

127 peaceful protests in the previous period (t=0). The expected total benefit by making peaceful protests at t=l would be as follows:p * SQ2 - CPROT.

This discounting factor may not be sufficient for members o f the ethnic group to make a decision to take up arms at t=l. In other words, the expected value achieved by peaceful protest may remain larger than that of violent rebellion: (p*SQ2 - CPROT) > (SQ2 - CREB). Then, members of the ethnic group will again choose peaceful protest at t= l despite some frustrations.

If peaceful protests continue for longer period without achieving their goals, however, the expected value of the change of status quo by peaceful protest will become smaller than the expected value o f the change of status quo by violent means at some moment, t=m: (pm*SQ2 - CPROT) < (SQ2 - CREB).

This effect of the discounting factor is illustrated in Figure 5.1.

Figure 5.1 Effect of the Discounting Factor Value SQ2-CPROT

SQ 2-CREB

SQ -\ — ------

0 m t

At such moment t=m, the accumulated frustrations among members of the ethnic group are so large that the organisers of collective action become disillusioned with peaceful protest as a means to successfully change an unfavourable status quo. As a result o f such disillusionment, some members of the ethnic group will take the other option, namely violent rebellion, even though the violent option is more costly than the peaceful option. Put simply, they choose a more costly option because they know from their own

128 experiences that the less costly option did not, and would not, work. In sum, the model discussed above suggests that grievances among members of the ethnic group would get larger when they keep failing to achieve their goals by peaceful means. Let us call this “dynamic grievances,” because it is assumed that the grievances increase over time, as members of the ethnic group continue their peaceful protest activities. One can thus draw the following hypothesis from this model:

Hypothesis 1: The longer members of an ethnic group have been involved in

peaceful protest activities, the higher the probability that some members

of the ethnic group decide to take up arms.

Of course, this discounting factor p cannot be calculated mathematically from the empirical analysis. In reality, this factorp may vary across ethnic groups or across individuals within the same ethnic group. Indeed, the case study in the following chapter shows that, in the case of Albanians in

Kosovo, some people gave up peaceful means earlier while other people kept sticking to the peaceful means, believing that they were the best option to successfully change the status quo. The purpose of the discussion above is simply to illustrate the problematic aspect of the assumption of structural determination model and to explain and incorporate the effect of the time dimension in a simplified manner.

(3) The Effect of the Type of Decision-Makers

As discussed in Chapter 1, the attention to the initial level of rebellion raises one question: do the low-intensity rebellion and large-scale rebellion have the same causes or not? Because the issue of initial level of rebellion has

129 not been addressed much in the existing literature, most analyses on the causes of the rebellion ignore the difference between the onset of low-intensity rebellion and high-intensity rebellion. Therefore, this chapter will attempt to examine whether the onsets of low-intensity rebellions and large-scale ones have the different causes or not.

How does this relate to the model discussed above? Recall the distinction between the two types of decision-makers - ethnic leaders and non-leaders. As was argued in Chapter 3, the key difference between these two types of decision-makers is the amount of resources available to them. One can argue that the amount of resources available to them will also affect the evaluation of the expected cost of making a rebellion.

As discussed in Chapter 3, non-leaders possess a limited amount of resources for the rebellion and often can organise only a low-intensity rebellion.

Due to the limited amount of resources, non-leaders seem more likely to perceive an option of violent rebellion as more costly: if one can only make a rebellion of a relatively low intensity, the balance of military powers will be in favour of the state authorities, and thus it will be more likely that the state authorities can crack down the rebellion and punish its participants harshly. For non-leaders, therefore, one can argue that it is more likely that the peaceful protest will be perceived as less costly than the violent rebellion, at least at the beginning. In other words, if they decide to take up arms, they will do so after trying to achieve their goals by peaceful means and realise that the peaceful option does not work, as discussed above.

For the ethnic leaders who can mobilise a large amount of resources, however, this may not be the case. Because they can mobilise a large amount of resources more easily, the procurement cost will be smaller than for the non-leaders. In addition, if they can mobilise a large amount of resources and

130 make a large-scale rebellion, the balance of powers between the state authorities and the rebels could be equal, or even in favour of the rebels.

Therefore, ethnic leaders may perceive that the violent rebellion is less costly than the peaceful protest. The ethnic leaders will then decide immediately to take up arms before trying to achieve their objectives by peaceful means.

This difference between leaders and non-leaders affects hypothesis 1 presented above. Note that hypothesis 1 is based on the assumption that members of the ethnic group decided to go to peaceful protests before they decide to take up arms. This means that, at least at the beginning of collective action, peaceful protest was perceived as less costly than violent rebellion by the organisers of collective action. The discussion above on the difference between ethnic leaders and non-leaders suggests that the validity of hypothesis

1 may be limited to the decisions made by non-leaders, because the assumption of hypothesis 1 may not be valid for the decisions made by ethnic leaders.

Because of a large amount of resources available to them, the ethnic leaders may decide to go to violent rebellions without making peaceful protests to achieve their objectives.

In order to check whether this argument is valid or not, this chapter will examine the occurrence of lower-intensity rebellion and large-scale rebellion separately.

2. Large-N Analysis

Having drawn a testable hypothesis based on the model, the present chapter now conducts large-N analyses to test it. This section firstly presents the research design and the operationalisation of variables. It will then present the empirical results.

131 (1) Research Design and Operationalisation

Dependent Variable

To test the hypothesis presented above, the dependent variable should capture the moment when some members of the ethnic group started to take up arms. For the dependent variable of the present analysis, therefore, a dummy variable (REBONSET hereafter) was created that takes a value of 1 when some members of an ethnic group are involved in violent rebellion activities (of whatever level) for the first time in the period of 1945-2000. An ordinal measure of “anti-regime rebellion” of the MAR dataset was used to create this dummy variable (as for the scale on which this variable is coded, see Appendix

1). In other words, this dummy variable (REBONSET) takes the value of 0 while this anti-rebellion index takes the value of 0. When this index takes a value more than 1 for the first time during the period from 1945 to 2000, the dummy variable takes the value of 1. This dummy variable thus captures the

“onset” of rebellion, i.e. the moment when some members of the ethnic group decided to take up arms against the state for the first time since 1945. The period after the “onset” of rebellion is given a missing value and will be excluded from the analysis. The unit of analysis is therefore ethnic group-year, and the dataset covers all ethnic groups and years from 1945-2000 except for the period after the initial onset of rebellion.

Some ethnic groups in the MAR dataset experienced more than two rebellions during the period from 1945 to 2000, separated by some continuous period of peace. Let us take an example of Tuareg in Mali in the dataset. The

“anti-regime rebellion” index for Tuareg takes the value of 6 for the first time in 1960 (this is the first year of the country as well) but becomes 0 in 1965

(note that, until 1984, the data is constructed based on 5-year periods). The peaceful period continues until 1989, and in 1990 the rebellion index takes the

132 value of 4 again and the rebellion continues until 1995. If one sticks to the operationalisation above, the dependent variable in all years after the first

“onset” in 1960 will have a missing value and the second onset of rebellion in

1990 will not be included in the analysis. In order to avoid this omission of later rebellions within the same ethnic group, the alternative operationalisation of the dependent variable will also be used, which takes the value of 1 whenever the rebellion starts (i.e. the rebellion index takes some value more than 1 and the value of the previous year is 0 or missing), the value of 0 when the ethnic group remains peaceful (i.e. when the “anti-regime rebellion” index takes the value of 0), and the missing value during the period of the continuation of the rebellion (The dummy variable based on this alternative operationalisation will be called REBONSET2 hereafter). The unit of analysis is therefore ethnic group-year, and the dataset covers all ethnic groups and years from 1945-2000 except for the period of continuation of rebellion. In the empirical analyses below, the results for the two operationalisations of the dependent variable will be compared.

Explanatory Variables

In order to test hypothesis 1, a variable “duration of peaceful protest activities” (PROTDUR hereafter) is created. This variable captures the number of years of continuous protest activities that members of the ethnic group had been engaged in. In order to create this variable, a variable of “peaceful protest” in the MAR dataset was used. This variable takes the values from 1 to

5 if members of the ethnic group are engaged in some kind of peaceful protest activities (such as demonstrations) and the value of 0 if no protest activity is reported for that year. The variable PROTDUR takes the value of 1 when some kind of peaceful protests (at whatever level) started in that year, and will keep

133 increasing by 1 every year as long as the peaceful protest activities continue.

When the peaceful protests stop in some year (i.e. the value of the “peaceful protest” index becomes 0), this variable takes the value of 0, and if the peaceful protests start again later, this variable again starts from 1. If hypothesis 1 is true, the probability of the onset of rebellion should increase as the duration of the peaceful protest activities increases. This variable PROTDUR is the variable of key theoretical interest in this chapter.

A number of control variables are included in the empirical analysis below in order to control for the structural factors discussed above. As for the motivating factors, the present analysis uses variables included in the MAR dataset to operationalise discrimination and autonomy grievances. Firstly, there are two variables in the MAR dataset - “political discrimination” (POLDIS) and “economic discrimination” (ECDIS) - that measures the role of public policy and social practice in maintaining or redressing inequalities. This variable is coded on a five-point scale, from 0 (no discrimination) to 4 (repressive policy / restrictive policies).1 The higher value in this variable means the higher level of discrimination against the ethnic group. One would thus expect that the probability of the onset of rebellion should increase as the level of discrimination gets higher. One disadvantage of using these variables is, however, that the data is available only for a limited period (from 1980 onwards) and the data is missing for a large share of the period covered by the present analysis. The inclusion of these two independent variables, therefore, will cause an omission of a majority of cases from the dataset (all cases from

1945 to 1979 are dropped from the analysis automatically). The estimation of the effect of other variables may be biased, therefore, by the inclusion of these

1 For the detailed description of the variables, see the codebook of the MAR data (see footnote 6 of Chapter 2 for the address of the website).

134 two variables that cause an omission of many cases in the dataset. In order to check this, the present analysis will conduct an analysis without these two variables to check the validity of estimation of the effect of the other independent variables.

As for the autonomy grievances, the “Index of Political Autonomy

Grievances” (AUTLOST) variable in the MAR dataset will be used as a proxy.

This variable is related to the loss of autonomy of the ethnic group and is a composite index for groups who have lost autonomy or undergone a transfer of control from one country to another. This is a continuous variable and takes a value ranging from 0 to 6, where the value of 0 stands for “no historical autonomy,” i.e. the absence of grievances related to the political autonomy, and the value of 6 stands for the highest degree of autonomy grievances.2 One would expect that the probability of the start of rebellion should increase as this variable AUTLOST takes a higher value.

The enabling conditions will be operationalised as follows. As for the presence of the highly authoritarian political regime, a dummy variable (DICT hereafter) was created, which captures the existence of a highly authoritarian regime, using the “polity2” variable in the PolitylV dataset.3 The polity2 variable in the PolitylV dataset shows the degree to which the political regime is autocratic or democratic, taking values ranging from -10 (strongly autocratic) to +10 (strongly democratic).4 The dummy variable DICT takes the

2 The variable is constructed by adding the weights for “Magnitude of Change” and “Group Status Prior to Change,” subtracting one, and dividing by the “Year-of Loss” weight. In other words, the coding of this variable considers (1) the degree of autonomy/independence the ethnic group had before the change, (2) the degree of loss after the change, and (3) the period of change. For the detailed description of the variable, see the codebook of the MAR data. 3 The PolitylV dataset is available at the website of the CIDCM, University of Maryland. See http://www.cidcm.umd.edu/inscr/politv/. 4 For detailed description of the variable and the coding procedure, see the “dataset user’s manual of the Polity IV project,” available online at: http://www.cidcm.umd.edu/ inscr/politv.

135 value of 1 when the polity2 variable of the country in which the ethnic group live takes the value between -5 and -10. If the argument in this chapter holds, this dummy variable must decrease the probability of the onset of ethnic rebellion, other conditions being equal.

As for the other enabling conditions, the present thesis again follows the operationalisation of Fearon and Laitin. The present analysis thus again uses their dummy variables of political instability and new state, which take the value of 1 if the country in which the ethnic group live had a three-or-greater change in the polity2 index in any of the three years prior to the year in question (INSTAB) and in the first and the second years of independence of the country in which the ethnic group live (NEWST). The present analysis also uses the GDP per capita (GDP) and the proportion of mountainous terrain

(MONT) of the country in which the ethnic group live, taken from Fearon and

Laitin dataset (see Chapter 3 for the details of these variables). As for the regional concentration of the ethnic group, the present analysis uses a dummy variable of the “regional concentration of the minority group” (REGCON) constructed by Fearon and Laitin. This variable takes the value of 0 when the minority group is either primarily urban or widely dispersed throughout the country, and a value of 1 if the group has a rural base somewhere in the country

(Fearon & Laitin 1999: 16). Fearon and Laitin (1999,2003) have found that all of these variables have statistically significant effect on the occurrence of civil war or large-scale ethnic violence.

Estimation Method

As for the estimation method, the present chapter uses the random-effects logit model. This model is used when observations occur in clusters and observations within a cluster (i.e. within the same ethnic group)

136 tend to be more alike than observations from different clusters, because ordinary analyses which assume that all observations are independent tend to produce invalid standard errors (Agresti 2002: 491). The random-effects logit model uses the cluster-level term in the model, which takes the same value for each observation in a cluster but different values for different cluster (this term is unobserved and treated as varying randomly among clusters). The advantage of this estimation method is that it takes into account the panel structure

(clustered structure) of the dataset and can estimate a more robust standard error for coefficients while it will retain the variability needed for estimating between-cluster effects of the explanatory variables in the model.5

Based on the construction of the random-effects logit model as a special case of the generalised linear mixed model for binary data (Agresti

2002: 492-502), the fitted model of the present analysis can be summarised as follows.

logit ij — /?o + U, + /?;POLDIS,y + frECDlSy + AUTLO ST y + /yDICTy +

/?5INSTAB/y + ^NEWST/y + /SyGDVij + /?sMONT,y + /^REGCONy +

AoPROTDURy

where

logit,y= log odds that ethnic group / experience the onset of rebellion in

the yearj

the random effect component {U,} are independent N (0, a 2) variates.

5 An alternative to the random-effects model for analysing clustered/panel data is to use the fixed-effects model. The severe disadvantage of this model, however, is that it will remove the source of variability needed for estimating between-cluster effects of the explanatory variables. If one is interested in the effects of the explanatory variables on the variance across clusters, therefore, one cannot use this model. See Agresti 2002: 496.

137 Effect of the Type of Decision-Makers

In order to check whether the low-intensity rebellion and large-scale rebellion have the same causes or not, one has to distinguish the onset of these two types of rebellions. If the argument presented above is true, the variable

PROTDUR will have a significant effect on the onset of low-intensity rebellion while it will not have a significant effect on the onset of large-scale rebellion.

In order to check the validity of this argument, the present analysis will conduct an additional analysis by splitting the REBONSET cases into two categories: LOWONSET, which takes the value of 1 when the REBONSET = 1 and the actual value of the anti-rebellion index is equal to or less than 5, and

HIGHONSET, which takes the value of 1 when the REBONSET = 1 and the actual value of the anti-rebellion index is equal to or more than 6. In other words, LOWONSET captures the onset of rebellions of lower initial intensity, while the HIGHONSET captures the onset of rebellions of higher initial intensity. If the argument of this chapter is right, the effect of the variable PROTDUR will have a significant effect on the onset of lower-intensity rebellion while it will not on the onset of higher-intensity rebellion. If it is wrong, on the other hand, there will not be such a difference in terms of the effect of PROTDUR on the onset of two types of rebellions.

(2) Empirical Results

The results of the empirical analysis with the first operationalisation of the dependent variable (REBONSET) are presented in Table 5.1. Model 1 is a model with all independent variables. Among ten independent variables, the effect of six variables is statistically significant at least at the 0.1 level, while the effect of the variables ECDIS, INSTAB, GDP and MONT is not statistically significant. The effects of all variables that are statistically

138 significant are in accordance with the theoretical expectations presented above.

The signs of the coefficients of PROTDUR, POLDIS, AUTLOST, NEWST and

REGCON are positive, which means that, other conditions being equal, the

probability that an ethnic rebellion will occur in that year increases as the

duration of the peaceful protest gets longer, as the level of political

discrimination and the level o f autonomy grievances get higher, when the

country is a newly-born state, and when the ethnic group is regionally

concentrated and has some rural basis. The sign of the coefficient of DICT is

negative, which means that the probability that an ethnic rebellion will occur in

that year decreases when the political regime in the country is strongly

authoritarian. The sign and the statistical significance of the coefficients hardly

vary when statistically insignificant variables are omitted from the analysis.6

Table 5.1 Rnndom-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: REBONSET) Model 1 Mocel 2 coef. std err coef. std. err. Political Discrimination 0.513** 0 2 . i . ■ Economic Discrimination -> > 0 196 IP t Lost Autonomy 0.535* 0318 0.889*** 0.179 Dictatorship -1 291** 0 624 -0.422* 0.241 Political Instability 0 239 0 469 0.395 0 253 New States 1.335* 0C69 1 823*** 0.374 Level of Economic Development -0.141 0 097 -0 08 0.057 Mountainous Terrain -0 144 0.199 -0.039 0 124

Regional Concentration 1 759** p 0 0 1 "86*** 0.443 Duration of Peaceful Protest 0064*** 0.023 0103*** 0 02 Constant -6 313*** 1.07 -6.740*** 0.652 N 2900 909- *p< 0.1; **p 0.05; *** p < 0.01

6 Results of the analysis are not shown here due to the lack of space, but are available from the author upon request. It is the same for the other analyses in this chapter.

139 As discussed above, the inclusion of POLDIS and ECDIS leads to the omission of a large number of cases in the dataset. In order to check whether this omission of a large number of cases biases the effect o f other explanatory variables, analyses were conducted without these two explanatory variables.

Model 2 is a model without POLDIS and ECDIS. Note that the number of cases increased from 2900 in Model 1 to 9097 in Model 2. This shows that more than 6000 cases (ethnic group-year) were omitted from the analyses due to the inclusion of POLDIS and ECDIS. The results presented in Model 2, however, show that this omission does not affect the results concerning the other explanatory variables very much. All explanatory variables that were found statistically significant in Model 1 remain statistically significant in

Model 2, and the signs of the coefficients of these variables also remain the same. In addition, all variables that were statistically insignificant in Model 1 also remain insignificant in Model 2. Again, the sign and the statistical significance of the coefficients hardly vary when statistically insignificant variables are omitted from the analysis.

Table 5.2 Random-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: REBONSET2) Model 3 Model 4 coef. std.en coef. std.en. Political Discrimination 0.245*** 0.093 Economic Discrimination -0.026 0.097 Lost Autonomy 0.306** 0.124 0 406*** 0088 Dictatorship -0.485** 0.235 -0.478*** 0.153 Political Instability 0.225 0 251 0.207 0.183 New States I 224** 0 498 1.947*** 0 276 Level of Economic Development -0070* 0037 -0.072** 0.029 Mountainous Tenain 0.09? 0092 0.062 0.071 Regional Concentration 0.980*** 0.313 0964*** 0.234 Duration ofPeacefi.il Protest 0.015* 0 008 0.033*** 0.006 Constant -5.241*** 0482 -5.168*** 0.323 N 4291 11276 * p < 0.1; ** p < 0.05, ***p<0 01

140 Table 5.2 presents the results of the analyses using the alternative operationalisation of the dependent variable (REBONSET2). Model 3 is a model using all explanatory variables. These results are mostly in line with the results presented in Table 5.1. All the explanatory variables that were found statistically significant in Model 1 remain statistically significant at least at the

0.1 level in Model 3, and the signs of the coefficients of these variables remain the same. The only difference from Table 5.1 is that the effect of GDP is now statistically significant at the 0.1 level. The sign of the coefficient of GDP is negative, which means that the probability that an ethnic rebellion occurs in that year decreases as the level of economic development gets higher. This result is in line with the theoretical expectations discussed above. These results remain intact when variables that were statistically insignificant in Model 3 are dropped from the analysis.

Model 4 is a model without POLDIS and ECDIS. The number of cases increases from roughly 4300 in Model 3 to more than 11000 in Model 4.

Again, however, the results are consistent with those of Model 3. All explanatory variables statistically significant in Model 3 remain significant in

Model 4, and the signs of the coefficients of these variables also remain the same. These results remain intact when explanatory variables that were statistically insignificant in Model 4 are dropped from the analysis. One can conclude, therefore, that the results presented above are not biased by the omission of a large number of cases caused by the inclusion of POLDIS and

ECDIS.

Going beyond the sign and significance of the coefficients, one can calculate the change of the odds based on the empirical results in order to visualise the impact of explanatory variables. To illustrate, let us see the change of the odds based on the results of Model 1 of Table 5.1. For example, the

141 coefficient of DICT is -1.291. This means that, other conditions being equal, the odds of occurrence of an ethnic rebellion will decrease by 73 percent if the political regime of the country is strongly authoritarian. The coefficient of

REGCON is 1.759, meaning that the odds of occurrence of an ethnic rebellion will be 5.8 times more (increase by 480 percent) if the ethnic group is regionally concentrated and has some rural base, other conditions being equal.

The coefficient of PROTDUR is 0.064. This means that, other conditions held constant, the odds of occurrence of an ethnic rebellion will increase by 6.6 percent as the duration of peaceful protest increases one year. Does this mean that the effect of PROTDUR is weaker than other variables, such as DICT or

REGCON? Not necessarily, since the variable PROTDUR can take a value from 0 to 55, while DICT and REGCON can only take either 0 or 1. While an effect of one-unit increase in the value of PROTDUR is smaller than that of

DICT or REGCON, an impact of increase by larger units in the value of

PROTDUR can be equal or even larger. For example, if the duration of peaceful protest increases from a minimum value (0) to a maximum value (55), the odds of occurrence of an ethnic rebellion will be 33 times larger (increase by 3200 percent), other conditions being equal. This increase of the odds is obviously much larger than that caused by DICT or REGCON. One can thus see that the effect of PROTDUR is no less significant than other variables such as DICT or REGCON, even though the latter variables have larger coefficients in Table 5.1.

Overall, therefore, one can conclude that the results of the empirical analyses in Table 5.1 and 5.2 support the key hypothesis of this chapter

(hypothesis 1) presented above. While some structural conditions (namely political instability and mountainous terrain) are not supported by any of the analyses conducted here, most of the structural conditions do have a significant

142 effect on the initial onset of ethnic rebellion, and their effects are all in line

with the findings in the existing literature.

Finally, let us conduct analyses separately for the occurrence of

rebellions of lower intensity and high intensity. The results o f the analyses

using LOWONSET as a dependent variable are presented in Table 5.3. Model 5

is a model with all explanatory variables. One can see from this table that the

sign and significance of coefficients remain the same as Model 1 o f Table 5.1,

except for DICT which became statistically insignificant. As for the effect of

PROTDUR, it remains statistically significant and its coefficient is positive.

These results are essentially the same when statistically insignificant explanatory variables are omitted from the analyses. When one excludes

POLDIS and ECDIS from the analysis, the results for explanatory variables again remain essentially intact except for DICT which again became statistically significant at the 0.1 level (see Model 6). These results hardly vary when statistically insignificant explanatory variables are omitted.

Table 5.3 Rundoni-Effect.s Logit Ann lysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: LOWONSET)

Mode . 5 Mode 16 Coef. std. err. coef std.en. Political Discrimination 0.320* 0 168 Economic Discrimination -0 104 0175 Lost Autonomy 0.497* 0.263 0.647*** 0.186 . i Dictatorship o .. 4- Q\ 0 513 -0474* 0 265 Political Instability 0 291 0456 0075 0 304 New States 1 517** 0.675 1.677*** 0 426 Level of Economic Development -0 061 0072 -0 018 0052 Mountainous Terrain -0 217 0 162 -0072 0 133 Regional Concentration 1.070** 0.52 1.458*** 0443 Duration of Peaceful Protest 0.060*** 0021 0 096*** 0.023 Constant -5 868*** 088 -6.580*** 0.7 N 2887 9057 * p • 0.1 • ** p 0 05. *** p < 0.01

143 The results of the analyses using HIGHONSET as a dependent variable are presented in Table 5.4. Model 7 is a model with all explanatory variables. Now one can see that the effects of explanatory variables are very different from Model 5. First of all, the effect of PROTDUR is now statistically insignificant. This result implies that the argument presented above on the effect of type of decision-makers could be true: while the duration of peaceful protests does increase the probability of occurrence of lower-intensity rebellion, it does not for the higher-intensity rebellion. This means that the organisers of collective action may immediately go to rebellion without going to peaceful protests when they have a sufficient amount of resources to organise a large-scale rebellion, as discussed above. Secondly, the effect of MONT is statistically significant at the 0.05 level. While there has been no empirical evidence so far for this variable, therefore, one can interpret this result as a partial empirical evidence for the argument on the effect of rough terrain.

Thirdly, the effect of ECDIS is now statistically significant. The sign of the coefficient is negative, however, contrary to the theoretical expectations presented above. In other words, this result suggests that the probability of occurrence of higher intensity rebellion decreases as the level of economic discrimination against the ethnic group gets higher. This contradictory result cannot be explained by the theoretical framework of the present thesis and thus remains a small puzzle. One ad-hoc answer to this puzzle could be that this factor may relate more to the economic resources of organisers of rebellion than to motivation of the organisers of rebellion. In any case, solving this puzzle is out of scope of the present thesis and this puzzle should be systematically addressed in future research.

144 Table 5.4 Random-Effects Logit Analysis of Determinants of the Onset of Rebellion, 1945-2000 (Dependent Variable: HIGHONSET)______Mode - Mode 18 coef. std.en coef. std.en Political Discrimination 3.220** 1 257 Economic Discrimination -1.997* 1.157 Lost Autonomy 0.885 1.022 : 317*#* 0 645 Dictatorship -4.363* 2 487 -0.997* 0.546 Political Instability -3.041 2.221 1 247** 0.485 New States 8 349 6.778 2.020** 0.844 Level of Economic Development -7.334*** n --> n -2.086*** 0.595 Mountainous Terrain 2.598** 1.116 -0 362 0.35 Regional Concentration 66.425*** 5.623 6.546*** 'I Duration of Peaceful Protest -0 024 0189 0.139** 0 061

Constant -S3 006 - -1464*** 3 467 N 2849 8970 *p 01. ** p 0.05; ***p<0.01

In the case of HIGHONSET, however, the results discussed above are not so definitive, because they do not remain intact when POLDIS and ECDIS are excluded from the analysis. Model 8 is a model without POLDIS and

ECDIS. Here, the effect of PROTDUR again becomes statistically significant at the 0.05 level, and the sign of the coefficient is positive in line with hypothesis 1. Therefore, this result suggests that the duration of peaceful protests may matter even for the onset of high-intensity rebellions. This implies that even ethnic leaders who can mobilise a large amount of resources for rebellion will still think that the peaceful protest is less costly than violent rebellion and try to achieve their goals by peaceful means before they go to rebellion. Secondly, the effect o f INSTAB is now statistically significant at the

0.05 level. The sign of the coefficient is positive, in line with the theoretical expectations of this chapter, meaning that the probability o f occurrence of higher-intensity rebellion increases if a country is in a transition. Note that the political instability is now the only variable that has not been empirically

145 supported so far. This result thus can be interpreted as a partial empirical evidence for the argument on the effect of political instability. These results remain intact when statistically insignificant variables are excluded from the analysis.

As for the effect of type of decision-makers presented above, therefore, the empirical evidence is mixed. On the one hand, when one includes POLDIS and ECDIS, the effect of PROTDUR becomes statistically insignificant for the onset of higher-intensity rebellions, while it remains statistically significant for the onset of lower-intensity rebellions. This suggests, as discussed above, that the argument of this chapter could be true: the organisers of collective action may immediately go to rebellion without going to peaceful protests, when they have a sufficient amount of resources to organise a large-scale rebellion. On the other hand, when one excludes POLDIS and ECDIS, the effect of PROTDUR is statistically significant for the onset of both lower-intensity and higher-intensity rebellions. This suggests that even ethnic leaders who can mobilise a large amount of resources will think that the peaceful protest is less costly than violent rebellion and try to achieve their goals by peaceful means before they go to rebellion. The empirical analyses here cannot provide a definitive answer. The following chapter, however, will show qualitatively that powerful ethnic leaders decided to go to rebellion without trying seriously to achieve their goals by peaceful means, because they thought that violent rebellions are not so costly to achieve their objectives.

Conclusion

This chapter attempted to explain the initial onset of ethnic rebellions.

In order to do so, this chapter presented a simple model that considers both

146 structural conditions and a dynamic aspect of grievances. This chapter has drawn a key hypothesis on the effect of dynamic grievances based on this model and then conducted a series of large-N analyses to test this hypothesis.

The large-N analyses empirically support this key hypothesis. As for the structural conditions, all control variables except for the political instability and the mountainous terrain have statistically significant effects on the initial onset of rebellion, as shown in Table 5.1 and Table 5.2. In addition, one can find partial empirical evidence for these two control variables in Table 5.3 and 5.4 where the empirical analyses are conducted separately for the onset of lower-intensity rebellions and higher-intensity ones.

The findings of this chapter have two theoretical implications. The first is the importance of the time dimension for the analysis of ethnic conflict.

As pointed out in Chapter 1, most theories in the existing literature remain structural and static, implicitly assuming that the decisions made by the members of ethnic group will be the same as far as the objective or structural conditions do not change. The present thesis does not have any intention to deny the importance of structural conditions: indeed, the empirical results of this chapter confirm the key findings in the exiting literature. This chapter pointed out, however, that they alone may not suffice and there may be a dynamic aspect of grievances: as members of the ethnic group continue to go to peaceful protests, they may get increasingly frustrated by the failure of peaceful protests to achieve their goals, even if the structural conditions remain the same. The results for PROTDUR are fairly strong in most of the analyses in this chapter and thus support this argument. This suggests that it is important to take into consideration this time-dimension and to move towards more dynamic models.

Secondly, this chapter has argued for the complementary nature of the

147 two types of structural conditions, namely the motivating conditions and the enabling conditions. As discussed in Chapter 1, some scholars argue that

“grievances do not matter” and that “opportunities”, or enabling conditions, suffice for the explanation of large-scale domestic conflict. However, the empirical results presented here support the argument for the complementary nature of these two types of structural conditions. In all empirical analyses in this chapter, at least one variable from both motivating conditions and enabling conditions remain statistically significant. As for the motivating conditions, political discrimination and loss of autonomy are found statistically significant in most of the analyses. As for the enabling conditions, the newness of the state and the regional concentration of the ethnic group are found statistically highly significant in almost all analyses. Overall, therefore, these findings support the argument for the complementary nature of the motivating conditions and the enabling conditions. Given the fact that “no study controlling for GDP and geographic concentration has shown that level of economic, cultural and political grievances can differentiate cases of high rebellion from cases of low or no rebellion” (Laitin 2002: 645), the findings of this chapter constitute an important contribution to the understanding of the causes of ethnic rebellion.

148 Chapter 6 Dynamic Grievances and Structural Conditions: Qualitative Analysis of the Ex-Yugoslav Countries

Introduction

This chapter conducts a case study of the four cases of rebellions in the ex-Yugoslav countries. It again aims to show that the theoretical arguments made in the previous chapter are based on some real-life cases and to make abstract arguments empirically more convincing. Chapter 5 argued that the dynamic aspect of grievances constitutes an important part of the explanation of the decisions by the organisers of rebellion to take up arms. This chapter will show that this dynamic aspect of grievances was an important factor which explains the timing of the decisions made by the KLA and the NLA to take up arms in Kosovo and Macedonia. This factor, however, was not so important in the case of Serbs in Croatia and Bosnia. This difference is due to the type of decision-maker as it was argued in Chapter 5: in the case of Serbs in Croatia and Bosnia, those who led the rebellion decided to take up arms earlier rather than later because they had a large amount of resources at their disposal and they thought that it was not so costly to take up arms. In analysing each case, this chapter also examines the effect of structural conditions discussed in

Chapter 5.

This chapter analyses each case systematically. It will firstly examine two cases of Albanian rebellions in Kosovo and Macedonia. It will explore the effect of the structural conditions in these cases and will show that, while some of these conditions are important for explaining the decisions to take up arms, the timing of the decisions made by the KLA and the NLA cannot be fully

149 explained only by these structural conditions. In both cases, the dynamic aspect of grievances was an important factor that explains the timing of the decisions to take up arms. It then analyses two cases of Serb rebellions in Croatia and

Bosnia, examining the effects of the structural conditions. This section examines how theoretical arguments in Chapter 5 fit (or do not fit) these two cases and will show that the newness of the Croatian and Bosnian state was an important factor that affected the perceived cost of taking up arms in the eyes of Serb leaders.

1. Albanians in Kosovo

This section firstly examines the structural conditions in Kosovo. It will be argued that these factors are important to explain the motivation of rebels, but they are not enough for explaining the timing of rebellion: all the structural conditions were present since the beginning of 1990s and thus they cannot explain why the rebellion started in 1996-97. There were two important factors that explain the timing - dynamic grievances and the availability of arms.

(1) Structural Conditions

Let us firstly examine the structural conditions. As for the motivating conditions, they do explain the motivation of the rebels. Firstly, Albanians in

Kosovo started experiencing harsh discriminatory policies implemented by the

Milosevic regime in the late 1980s, as discussed in Chapter 4. The discrimination against Albanians was institutionalised by a series of laws adopted in the Parliament of Serbia. Albanians thus experienced a serious level of discrimination, such as expulsion of Albanians from the Kosovo government,

150 dismissal of Albanian workers in factories and other workplaces, and prohibition of education in the Albanian language. Secondly, Albanians enjoyed a high degree of autonomy under the 1974 constitutional regime in the socialist Yugoslavia, and the Milosevic regime effectively annulled the autonomous status of Kosovo. In other words, the Albanians in Kosovo experienced a loss of autonomy they once enjoyed under the socialist regime.

One can thus argue that the grievances were very high among Albanians. For example, Ballentine argues that the main drivers of conflict in Kosovo were

“the grievances and insecurity bred by the systematic exclusion of ethnic minorities from political power and an equitable share of economic opportunities and benefits.”1 As discussed in Chapter 4, these grievances motivated Albanians to make a series of peaceful activities to seek independence of Kosovo in the 1990s. Those who led the armed rebellion of the KLA shared the purpose, while they disagreed on the means to achieve it.

For example, Kelmendi told as follows: “if we fail to achieve our goals by political means, we are entitled to resort to armed resistance. That’s why we took up arms.”2

As for the enabling conditions, their effects are mixed. On the one hand, the Albanians were under a quite harsh authoritarian regime (polity2 value is -5 from 1988 to 1992, -7 from 1993 to 1996, -6 from 1997 to 1999);

Serbia (Yugoslavia) was not a newly independent state; Serbia (Yugoslavia) was not experiencing a political transition before 1999 (according to the operationalisation adopted by Fearon and Laitin). Therefore, one can argue that the enabling conditions were not so favourable for the rebellion with regard to the capabilities of the state authorities. Indeed, as discussed in Chapter 4,

1 Ballentine 2003: 260. See also Yannis 2003 for more detailed analysis of the case of Kosovo. 2 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30.

151 Albanian political leaders had chosen a non-violent strategy partly because they thought that the Serbian military forces were too strong to fight against.3

These factors thus explain thelack of rebellion until 1996 in Kosovo, but they cannot explain the occurrence of rebellion, since these factors did not change when the rebellion started in Kosovo.

On the other hand, however, other enabling conditions do seem to explain the case of Kosovo. Firstly, the level of economic development of

Yugoslavia was not high, if not very low like Sub-Saharan countries, at the beginning of the 1990s (somewhere around the mean value of 161 countries in the dataset of Fearon and Laitin), and furthermore, Kosovo was the least developed region in the former Yugoslavia.4 Secondly, Albanians are concentrated in Kosovo and they do have a rural base. Indeed, the KLA appeared from the rural part of Kosovo, and the members of the KLA were mostly from rural areas. For example, Shemsi Syla, who was then a zone deputy commander of the KLA in Karadak area, confirmed that most of the

KLA members came from rural areas and the military operations were conducted mainly in rural areas “because the configuration of the terrain was more adaptable.” 5 Thirdly, the geographic terrain of Yugoslavia is also relatively mountainous (slightly below the mean in the dataset of Fearon and

Laitin) and the terrain in Kosovo is more mountainous compared to the northern part of Serbia. Some argue that the mountainous terrain in the border area between Kosovo and Albania helped the guerrilla insurgency of the KLA.

One report points out, for example, that “in the wild and remote Albanian highlands it is virtually impossible to detect all the ancient mule tracks that

See the quote of Rugova’s comment in Chapter 4 (p. 106), 4 For example, GDP per capita in Kosovo was only 27% of the average of Yugoslavia and one-eighth of Slovenia in 1988. See Plestina 1992: 180-181. 5 Interview with Shemsi Syla, Prishtina, 2006/05/10.

152 snake their way through otherwise impenetrable mountains” and thus it was extremely difficult to police the smuggling of the arms into Kosovo (ICG

1998b: 13). General Bozidar Delic from the Yugoslav Army (Vojska

Jugoslavije, VJ hereafter), who was a commander of the 549th Motorised

Brigade stationed in Kosovo during the period of conflict, also emphasised the difficulty in defending the border area due to the mountainous terrain 6

(2) Dynamic Grievances and the Availability of Arms

While some of the structural conditions do seem to matter for the onset of rebellion in Kosovo, they alone cannot explain the timing of rebellion.

As discussed in Chapter 4, the rebellion organised by the KLA started in 1996-97. There was, however, almost no change in the structural conditions in or around this particular year. The level of discrimination and the loss of autonomy had been constant since the late 1980s and the beginning of the

1990s. There was also no change in the authoritarian nature of the Milosevid regime since the beginning of the 1990s. The economy in Kosovo had also been collapsed since the beginning of the 1990s. Needless to say, the geographic terrain does not change in such a short period. In a word, therefore, there seems to be no change in the structural conditions that could explain the timing of the onset of rebellion in Kosovo.

What then explains the timing of the decision to take up arms by the

KLA? In the case of Kosovo, there were two crucial factors that determined the timing of the start of rebellion by the KLA. The first factor was what was conceptualised as dynamic grievances in Chapter 5: the growing disillusionment and frustration among the ethnic group that decreased the support for the non-violent strategy and increased the support for the violent

6 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 9339 (2005/06/22).

153 means. The second factor was the availability of arms.

The first factor was the growing disillusionment among Albanians on the non-violent strategies. In the late 1980s and the early 1990s, an overwhelming majority of Kosovo Albanians supported the non-violent resistance led by the LDK. In May 1992, for example, the LDK won the overwhelming majority of votes (76.44%) and Rugova won 99.5 percent of the votes as the only candidate for president. As Clark (2000: 83) pointed out, the

LDK was “more of a national movement than one party among several.”

Especially after the 1992 elections, the LDK behaved as “the only true representative of the national interest of Kosovo - as much at a village level as the international level” (Clark 2000: 118).

The Albanian political leaders in Kosovo struggled to achieve the independence of Kosovo by peaceful means, especially by appealing to major powers to recognise the independence of Kosovo. Their efforts, however, could not yield any results. While many international conferences were held on the Yugoslav crisis since 1991 and the Albanian leaders attempted to put the issue of Kosovo on the agenda, their efforts yielded no results. In 1994 and 1995, so-called “K+K plan” was popular among Kosovo Albanians, according to which Kosovo Albanians would seek what the Krajina Serbs in Croatia would receive (Simic 2000: 96), but this plan became unrealistic when the self-proclaimed state of Krajina Serbs collapsed in 1995. The Albanian leaders lobbied the US government to solve the issue of Kosovo at the Dayton peace conference,7 but their efforts failed again. In September 1996, Rugova made an agreement with Milosevic on the normalisation of education, brokered by a

Catholic organisation Comunita di Sant’Egidio,8 but this agreement was never

7 Interview with Edita Tahiri, Prishtina, 2006/05/09. 8 For the details of the agreement, see Simi6 2000: 99-100; Troebst 1998: 81-83.

154 implemented.

The successive failures to achieve their goals by peaceful means led to the increasing frustrations among Albanians in Kosovo and the growing disillusionment with the non-violent resistance. Already by 1993, there was a growing resentment at international negotiators for not according the Kosovo

Albanians full status at Geneva, and this resentment had begun to shift the balance within Rugova's movement toward radical militants who preached a military solution (Woodward 1995: 359). In March 1994, Bukoshi told that the

“Kosova government’s pacifist approach was losing credibility within the population” (Clark 2000: 118). The Dayton agreement in 1995 was the turning point in this regard. The Dayton agreement itself was “an extraordinary trauma for the Kosovo Albanians” because it rewarded violent actions taken by Serbs with the recognition of Republika Srpska, which “confirmed to them in the most dramatic and humiliating way that Rugova’s policy of passive resistance had failed” (Judah 2000a: 124-125). Simic (2000: 54) observes that the failure of the LDK to put the question of Kosovo on the agenda at the Dayton peace conference led to the division within the LDK cadres and the radicalisation of

Albanian youths and Albanian national movements. Florin Krasniqi, an

Albanian immigrant who organised the financial assistance for the KLA from the US, told as follows: “Dayton made me realize that fighting was the only way. Peaceful resistance brought us nothing. Every other disgruntled minority in Yugoslavia...fought for independence and got it. The Albanians took the peaceful road and were ignored” (Sullivan 2004: 3).

Actions taken by the international actors after Dayton furthered the disillusionment among Albanians. The UN embargo imposed on Yugoslavia was lifted, and the EU states officially recognised the Federal Republic of

155 Yugoslavia.9 Bonn and Belgrade made an agreement and Germany returned

100,000 Albanian refugees to Yugoslavia, and “it appeared as if the last international means to put pressure on Belgrade were also lost” (PetriC & Pihler

2002: 74). Thus the IICK (2000: 59) concluded, for example, that the international community sent a message that “Kosovo was definitely off the current international agenda” and this demoralised and weakened the non-violent movement in Kosovo which “felt betrayed by the international community and began to doubt the effectiveness of its own tactics.”

Some argue that the Albanian political leaders led by Rugova and the

LDK were also responsible for this disillusionment, because they gave an impression that the issue of Kosovo would be resolved in the peace process in Croatia and Bosnia. For example, a prominent Albanian intellectual in Kosovo,

Veton Surroi, told that “Rugova personally, and his information apparatus, gave the impression that the question of the independence of Kosova was a matter of days, and that the entire international community was dealing on a daily basis with the question of Kosova. Over a critical period, for instance, prior to the

Dayton talks, the Democratic League of Kosovo apparatus was basically going about saying the question of Kosova would be solved, resolved at Dayton.”10

Rugova’s influence and authority decreased further by the failure of the agreement with Milosevic in 1996. The failure of this agreement, which offered a short spate of hope for improvement, greatly increased people’s frustration, and Rugova was seen as unwise to have trusted that Milosevic would follow through on anything that mattered to the Albanians (Trix 2005).

Trix (2005: 327) points out as follows: “Throughout 1997, the

9 Note, however, that the US maintained so-called “outer wall sanctions” and blocked Yugoslavia’s return to the World Bank and the International Monetary Fund, citing the issue of Kosovo as one reason. 10 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 3373-3375 (2002/04/18).

156 nonimplementation of the St. Egidio Education Agreement festered, giving fuel to the cynics and the extremists. If there could be no agreement on something as basic as school buildings, how could there ever be trust for larger issues?”

Disillusionment with the non-violent resistance policy started to spread among Albanians both within and outside Kosovo. Within Kosovo, for example, students in Prishtina waged protests in August 1997 against the failure of the Serbian government to implement the 1996 agreement on education. What is important here is the fact that they ignored Rugova when he asked them to stop it (Judah 2000a: 135-136). The division grew among the

LDK members as well and some members who wanted more radical actions left the party (Judah 2000a: 136). Hydajet Hyseni, then vice-president of the

LDK who led the party’s “radical” wing, said in autumn 1997 that the time had come to abandon Rugova's policy of non-confrontation, saying “We cannot be satisfied with our results or the attitude of the international community... The militant tendency among Albanians is the consequence of unproductive policies in Kosovo.”11 According to Edita Tahiri, former Foreign Minister of the govemment-in-exile of Republic of Kosova, “almost everybody in the presidency of the LDK was for the uprising, including Fehmi Agani, by the end of 1997. Only Rugova was against the violent uprising.” 12 This disillusionment with the non-violent strategies among local elite and ordinary

Albanians provided an opportunity for recruitment for the KLA.

Disillusionment among Albanians also grew outside Kosovo, which led not only to the increased recruitment abroad but also to the increasing flow of money into hard-liners. Salihu and other members of the LPK could tour the clubs and meetings of Kosovo immigrants and took advantage of the

11 Financial Times, 1997/11 /04. 12 Interview with Edita Tahiri, Prishtina, 2006/05/09.

157 disappointment among them (Judah 2000b: 320). Some Albanian immigrants started switching their support from the LDK to the “Homeland Calling” fund set up by the KLA (IICK 2000: 52). In the case of Kosovo, therefore, the growing disillusionment with the non-violent policy and the growing frustrations among the Albanians led to the increased support for the hard-liners, which made it possible for the hard-liners to acquire and mobilise more human and financial resources for their objective.

The second important factor was the availability of arms. For hard-liners, even when they were willing to start uprising, “the simple fact remained that there was no way to obtain and import large amount of weaponry” (Judah 2000a: 127). This problem was solved by the collapse of the neighbouring country, Albania. When the pyramid schemes of the investment collapsed in spring 1997, the Albanian government lost control and country fell into chaos. With the military dissolved, police running away and the army depots thrown open, hundreds of thousands of Kalashnikovs and ammunition were looted by local Albanians. The KLA began buying these guns and ammunition. While the exact number of arms that flowed into Kosovo is unknown, it is suggested that the amount was indeed huge. According to some report, for example, about 750,000 weapons were stolen from military depots and many of them ended up in the hands of the KLA (ICG 1998b: ii).

According to General Obrad Stevanovid from the VJ, the number of weapons seized by the Serbian authorities by 20 June 1999 amounted to 1045 hand-held rocket launchers, total 8320 pieces of machine-guns, semi-automatic rifles, automatic rifles and similar rifles, 360 pistols, 4224 mines and other explosive devices, and 723,531 ammunition of different types.13 Without this massive amount of arms suddenly available for the KLA, the uprising would have been

13 ICTY transcript of case IT-02-54 (Slobodan MiloSevic): 39574 (2005/05/17).

158 technically impossible despite the increasing willingness of the KLA members to resort to violent means.

2. Albanians in Macedonia

The previous section pointed out that the dynamic aspect of grievances was an important factor for the timing of the decision to take up arms in

Kosovo. The same point can be made in the case of Albanians in Macedonia: while the structural conditions explain the motivation of the rebels, they cannot explain the timing of the onset of rebellion because there was no change in the structural conditions when the rebellion of the NLA started in 2000. This section will show that the dynamic grievances and the availability of arms again explain the timing of rebellion.

(1) Structural Conditions Let us firstly examine the structural conditions in Macedonia. As for the motivating conditions, Albanians in Macedonia did feel that they were discriminated against by the Macedonian authorities on various issues, such as the recognition of citizenship, the lack of university education in the Albanian language, prohibition of the usage of Albanian symbols in official places, and the lack of equal opportunity for employment especially in the public sector

(see Chapter 4). While the level of discrimination is certainly not as high as in

Kosovo, one can argue that there was a certain level of discrimination against

Albanians in Macedonia as well. As for the loss of autonomy, on the other hand,

Albanians in Macedonia had never enjoyed autonomy and thus one can conclude that there was no experience of loss of autonomy or independence among the Albanians in Macedonia.

159 Even though the level of discrimination was not comparable to

Kosovo, these discriminations did indeed motivate Albanians to take up arms.

The fact that these grievances motivated the rebels can be seen in their declared demands. By examining the published communiques and other documents,

Rusi shows that demands of the NLA were highly consistent. Among others, the NLA demanded that the constitution of Macedonia should be changed so that the following six elements are sanctioned, namely (1) Macedonia will be a state of two peoples, Macedonians and Albanians, (2) Albanian becomes an official language, (3) each community will be free to use its own national symbols, (4) discrimination in the economy and in the state administration will be eliminated, (5) discrimination in the political system will be eliminated, and (6) all political prisoners will be freed and the right to return will be recognised to all the people persecuted for their political beliefs and the people who fled

Macedonia (Rusi 2004: 2-3). Most of these demands were consistent with the demands made by Albanian political leaders in Macedonia throughout the 1990s. Even ArbSn Xhaferi, an Albanian political leader who criticised the

NLA and called for the halt of violence, said that “the demands of the fighters are the same as ours” (Phillips 2004: 120).

Of course, one might argue that the rebels often justify the recourse to violence by grievances and thus it is too naive to believe their words. It is not just their words, however, but also their actions that show that they were mainly motivated by these grievances: when their demands were mostly met in the Ohrid agreement in August 2001, most of the NLA fighters stopped fighting and accepted the disarmament. If the rebels were motivated by other causes, such as greed, and they used grievances as a mere pretext of using violence, they would not have stopped violence even when the Ohrid agreement was made. Since it was a result of the concessions, it would have

160 been easy for the NLA to claim that the Ohrid agreement does not fully satisfy their demands and to continue their armed struggle.

As for the enabling conditions, their effects are again mixed. On the one hand, Macedonia became independent in 1992 and its military capacities were extremely weak because of the withdrawal of the JNA from Macedonia, but no rebellion occurred during the first several years since independence.

Unlike Serbia, its political regime was not a strongly authoritarian regime

(polity2 is 6 for all years from 1993 to 2000) and it did not experience any political instability (as operationalised by Fearon and Laitin) since 1993. As for the geographic terrain, 8.4% of the terrain is estimated to be mountainous in

Macedonia, which is below the mean in the Fearon and Laitin Dataset. These factors, therefore, do not explain the case of Macedonia very well. On the other hand, however, the level of economic development of Macedonia has not been very high (its value in 1993 is below the mean of the Fearon and Laitin dataset), and the Macedonian economy experienced a serious hardship due to the transition from the socialist to the capitalist economy, the economic sanctions imposed on Yugoslavia by the international community and the economic sanctions on Macedonia imposed by Greece. The Albanians are concentrated in the western part of Macedonia and they have a rural base. The level of economic development and the regional concentration of the ethnic group, therefore, may partly explain the onset of rebellion in Macedonia.

In any case, what the structural conditions cannot explain is the timing of the rebellion. There was almost no change in the structural conditions discussed above in Macedonia. The level of discrimination basically remained the same since the beginning of the 1990s, showing no sign of either drastic improvement or serious deterioration. The level of economic development was indeed rising since 1996 (see Table 6.1). Macedonia did not experience any

161 political instability and Macedonia cannot be regarded as a newly-born country in 2001, at least according to the operationalisation of Fearon and Laitin. The structural conditions, therefore, cannot explain why the rebellion of the NLA started in 2001.

Table 6.1 GDP Real Growth Rates in Macedonia, 1991-2000

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

-6.2 -6.6 -7.5 -1.8 -1.1 1.2 1.4 3.4 4.3 4.5

Source: The report of the statistical office of Macedonia, 3.1.5.06, 07/12/2005, http://www.stat. gov.mk/pdf/1-2005/3.1.5.06.pdf

(2) Dynamic Grievances and the Availability of Arms What explains the timing of the rebellion in Macedonia? Two factors pointed out above in the case of Kosovo also explain the timing of the rebellion in Macedonia, namely (1) the growing frustrations among Albanians (dynamic grievances) and (2) the availability of arms.

The first is the accumulation of the frustrations among some Albanians caused by the constant delay of the reforms to address their demands. Let us take an example of Tetovo University. Albanians in Macedonia demanded the recognition of higher education in the Albanian language for many years, and the university started functioning illegally in Tetovo as early as in 1995.

Albanians demanded that the Macedonian government officially recognise the degrees from this university. In the summer 1999, the Macedonian government started to consider the official recognition of the university in return for the fact that Albanians in Macedonia remained calm despite the escalation of Kosovo

162 conflict and the influx of Albanian refugees in Macedonia.14 However, the recognition did not materialise due to the opposition from Macedonian parties and demonstrations by ethnic Macedonian students. The other long-held demands by Albanians were in the similar situation: despite the continuous participation of Albanian parties in the government, demands made by

Albanians were not met and their realisation was constantly delayed.

This constant delay of the reforms led to the growing disillusionment with the established Albanian political elite. In the case of Macedonia, it was the change of the party composition in 1998 that fostered the disillusionment with the Albanian political elite. The participation of the Democratic Party of

Albanians (Demokratska partija na albancite, DPA here after) in the coalition government after the 1998 elections seemed to promise rapid reforms, but indeed little change was forthcoming thereafter (Pettifer 2001). As a result, according to Pettifer (2001: 10), a situation was created “where a classic conflict scenario has unfolded, where the promise of reform after a long period of repression has not been met, and as a result more radical leaders, in this case

Mr Ali Ahmeti and the NLA, have begun to replace constitutionalist leadership.” The Albanians who were sympathetic towards the NLA stated that they lost patience because in the last 10 years the Macedonian authorities did nothing regarding their demands.15 One Albanian said: “for 10 years, we tried to solve this with politics, but the Macedonians weren’t interested. Now there’s no alternative.”16

As the rebels were motivated by the disillusionment with the Albanian political leadership, an armed uprising was an action not only against the state

14 See e.g., RFE/RL Balkan Report, 3-28, 1999/07/20; RFE/RL Balkan Report, 3-30, 1999/08/03. 15 AIM Press, 2001/03/22. 16 Daily Telegraph, 2001/03/26.

163 authorities but also against the Albanian party elite in Macedonia. For example, in a letter to Kofi Annan and other international figures, the NLA accused, albeit vaguely, the Albanian political parties in Macedonia for the failure to bring about the positive change for Albanians (Rusi 2004: 6). Those who interviewed the NLA members reported that their interviewees were consistently critical of Albanian politicians in Macedonia. When Ali Ahmeti talked to foreigners, for example, he bluntly described the Albanian politicians as “looking after their own interests” (Rusi 2004: 7).

Disappointment with the Albanian parties was caused not only by their ineffectiveness in bringing about positive changes but also by corruption.

According to Glenny, the DPA apparently argued for increased representation of Albanians but actually expended most of its energy in enriching itself thanks to the symbiotic relationship of corruption that it developed with its coalition partner, the Internal Macedonian Revolutionary Organization - Democratic

Party for Macedonian National Unity ( Vnatresna Makedonska revolucionema organizacija - Demokratska partija za Makedonsko nacionalno edinstvo,

VMRO-DPMNE hereafter).17 The Prime Minister Georgievski and his colleagues became deeply implicated in the process of “asset stripping” whereby foreign companies buy up utilities as part of the privatisation process.

The VMRO-DPMNE could not do this without the collusion of the DPA and the deputy President of the DPA, Menduh Thaci, became their ideal partner.

Glenny (2002) thus argues that the rebellion of the NLA was aimed “clearly at the DPA and Thaci's greed” that had excluded the majority of Albanians from the fruits of corruption. According to one report, many local Albanians claimed that the emergence of the NLA was the result of “corruptness of DPA

17 The argument in this paragraph is based on Glenny 2002.

164 representatives” as well as “unkept promises.”18

The second important factor that explains the timing of the NLA rebellion is the availability of arms and experienced personnel. Here, the end of the Kosovo conflict in 1999 was of critical importance. With the end of the

Kosovo conflict, Albanians from Macedonia who participated in the rebellion in Kosovo returned to Macedonia. For example, two of the principal founders of the NLA, Ali Ahmeti and Amrush Xhemajli, were founding members of the

KLA (ICG 2001: 8). Gezim Ostreni, chief of the NLA Headquarters, was an experienced middle aged soldier with an ex-JNA background who until April

2001 was a deputy commander of the Kosovo Protection Corps (KPC hereafter) that was set up after the war in Kosovo and incorporated many former KLA soldiers (Rusi 2004: 13; Pettifer 2001: 12). Other leaders of the

NLA include Xhavit Hasani, Skender Habibi and Emms Dzemali, long-time leaders of the KLA secret services (ICG 2001: Appendix A). According to

Ordanoski (2004: 20), among the NLA fighters were a few hundred so-called

“dogs of war,” who had gained experience at the fronts in Croatia, Bosnia and

Kosovo. Most of them were Albanians from Kosovo, former or active members of the KPC, and they were highly mobile and equipped with sophisticated western arms. According to Rusi (2004: 12), the NLA members who started operations in Macedonia were recruited from Ramush Haradinaj’s former fighters in the territory around Prizren.

The degree to which the rebellion in Macedonia was “imported” from

Kosovo remains, however, contentious. On the one hand, some scholars argue that the NLA came from Kosovo, which is understandable given the profile of the NLA leaders. For example, Janev (2003: 314) states that the NLA came

“from training camps in Kosovo and from the ranks of the never-disbanded

18 AIM Press, 2001/03/22.

165 Kosovo Liberation Army.” On the other hand, some emphasise the local origins of the NLA. The ICG (2001: 9) points out that people who came from Kosovo were “so small and fragmented that they cannot plausibly be credited with starting single-handedly such an effective insurrection.” Roudometof (2002:

215) argues that it became apparent that the NLA was not a small gang of criminals or some freebooters who sneaked from Kosovo, and the NLA rebellion was well organised and supported by many Albanians in Macedonia.

Pettifer (2004) also emphasises the difference between the KLA and the NLA:

“compared to the KLA, the NLA was quite a difficult organisation to join: it was much better armed and trained and had a different internal paramilitary culture and modus operandi.” As pointed out in Chapter 4, major political leaders in Kosovo, such as Tha^i and Mahmuti, argued that it was wrong to make another rebellion in Macedonia.19 According to one report, no one in

Kosovo denied that there could be Kosovo Albanians participating in the armed movement in Macedonia, but “the idea that the movement was sent in from Kosovo is firmly rejected.”20 The neutral judgement is difficult since parties to the conflict hold completely opposing views: the Macedonian authorities considered the rebels to be from Kosovo or local criminals,21 while the NLA claims that all of their soldiers were from Macedonia and they have a “strict policy not to accept any soldiers from other Albanian lands in the region” (Rusi

2004: 6). In any case, however, one can safely conclude that the availability of arms and experienced personnel for the rebellion is an important factor that explains the timing of rebellion in Macedonia.

19 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30. See Chapter 4, footnote 38 (p. 115). 20 AIM Press, 2001/03/22. 21 AIM Press, 2001/3/12. See also Roudometof2002: 215; Qrdanoski 2004: 17.

166 3. Serbs in Croatia

This section examines the effects of structural conditions in the case of

Serbs in Croatia. As for the motivating conditions, the key factors that motivated SDS politicians to take up arms were discrimination against Serbs after the transition to the multi-party system and the moves for independence from the socialist Yugoslavia. Unlike the cases of Albanian rebellions in

Kosovo and Macedonia, however, the dynamic grievances do not explain the case of Serb rebellion well, since the Serb elites in Croatia had chosen to take up arms sooner rather than later without trying to achieve their goal by peaceful means. By examining the effects of enabling conditions, this section will show that it was the newness of the Croatian state that lowered the perceived cost of rebellion in the eyes of the Serb leaders in Croatia, even though it turned out to be a miscalculation.

(1) Motivating Conditions Let us firstly examine the effects of motivating conditions. As for discrimination, the Serb elite and some analysts argue that the Serbs indeed experienced discrimination after the multi-party elections in Croatia in 1990 and it motivated the Serbs to take up arms.22 For example, after the HDZ took power, Serb police officers started to be dismissed and were replaced by Croats

(Silber & Little 1996: 98-99). In a new Croatian constitution adopted in

December 1990, “the millennial identity of the Croatian nation and the continuity of its statehood” were emphasised at the beginning, and the

Republic of Croatia was defined as “the national state of the Croatian nation.”23

22 For such arguments made by scholars, see, e.g., Cohen 1993: 126-135; Silber & Little 1996: 98; Woodward 1995: 134; Glenny 1996: 12. 23 For the original text of this constitution, see Marko & Boric 1991: 437-448. The text is

167 There was no specific provision on minority rights, and the Croatian language and the Latin script became the official language and script. In sum, it was obvious that Croats would be in the hegemonic position in the new state. This dominance of Croats after the transition to the multi-party system and the increased level of discrimination against Serbs were clearly one of the motives of the SDS politicians for taking up arms.

It was not only discrimination, however, but also the moves for independence taken by the Croatian authorities that motivated the Serb leaders to take up arms. The new government in Croatia started its moves for independence immediately after the elections. In June 1990, for example, the then vice-president of the Croatian Parliament Vladimir Seks mentioned the idea of “confederation” and revealed the idea that each republic should have its own embassy and military.24 In October 1990, the governments of Slovenia and Croatia announced a plan to transform Yugoslavia into a confederation of six sovereign independent republics.25 In February 1991, the Croatian Parliament approved an amendment of the constitution of the republic that stipulated that the republican constitution precede the federal constitution.26 In

May 1991, the Croatian government organised a referendum on independence.

These moves for independence dissatisfied the Serbs in Croatia. In July 1990, for example, Babi6 stated in an interview that they would seek regional autonomy if Croatia secedes from Yugoslavia.27 When Babic announced the establishment of the SAO Krajina in December 1990, he criticised the Tudjman government, saying “Croats are attempting to make their own state by a

also available from the website of the Official Gazette of Croatia. See, http://www.nn.hr/ sluzbeni-list/sluzbeni/index. asp 24 Politika, 1990/06/29. 25 Borba, 1990/10/08. 26 Borba 1991/02/21. 27 Politika, 1990/07/10.

168 majority of the parliament, hegemony, and the tyranny of majority, while the most important state organisation for Serbs in Croatia is Yugoslavia.” One can argue, therefore, that the moves for Croatian independence motivated the

Serb leaders to decide to take up arms.

(2) Enabling Conditions

As for the enabling conditions, their effects are again mixed. Croatia was a relatively developed country (GDP per capita in 1995 was above the mean of the Fearon and Laitin dataset) and more developed within the socialist

Yugoslavia, ranking 2nd after Slovenia among the six republics and two autonomous provinces (Plestina 1992: 180-181). On the other hand, the Knin region which became the regional basis for the Serb rebellion was characterised by weak economic development and a rural population (Rimac et al 1992; Sterc 1992). Croatia’s geographic terrain is not so mountainous (only

3.6% of the terrain is mountainous in Croatia, which is below the mean of the Fearon and Laitin dataset) but the Knin area is more mountainous than other regions in Croatia. Croatia experienced a transition from a one-party regime to the multi-party regime in 1990, even though this transition is not captured by the polity2 dataset since Croatia enters the Polity IV dataset only in 1991 (note that, if one assume the same value for all republics in Yugoslavia, the value of the polity2 variable for Croatia (Yugoslavia) is -5 since 1980 to 1990 and it is

-3 for Croatia in 1991, thus the variable of political instability would not take a value of 1 in any case). The Serbs had some rural base where they were concentrated and the rebellion was indeed organised in these rural areas.

One important variable among the enabling conditions is the newness of the Croatian state. Fearon and Laitin argue that the newness of the state

28 Politika, 1990/12/22.

169 implies a weakness of the new state authorities and increases the probability that the civil war occurs in that country. This logic can be observed in Croatia as well. The republics did not have their own army during the period of the socialist Yugoslavia29 and Croatia started building up its own army from scratch only after the HDZ won the elections in 1990. In addition, as discussed in Chapter 4, the Serb leaders had a large amount of resources for the rebellion at their disposal. This (perceived) weakness of the Croatian military and the strength of the Serb military forces affected the decisions by the SDS leaders to take up arms immediately after the declaration of the independence by the

Croatian authorities. For example, Milan Babic thought that a war against

Croatia would be easy and urged his more moderate colleagues not to negotiate with the Croatian government: according to Jovan Opafcic, one of the founders of the SDS in Croatia, Babic was arguing in the summer of 1991 that “it will take us only fifteen days to deal with the Croats by force of arms.”30 Such a belief was clearly affected by the optimism held by the JNA generals: in describing the plan of the JNA against Croatia in his memoir, Kadijevid (1993:

135-136) stated that it was assumed that “it would require 10-15 days” for the mobilisation, the preparation of the mobilised units and their action against

Croatia. Babic claims that it was Milosevic who assured him that the Serbs in

Croatia will be protected by the JNA: according to Babic, Milosevic told him

“the JNA will protect you” when he expressed his fears and anxieties as to

29 While the TO forces were under the leaderships of the republic and autonomous provinces in peacetime until 1988, they were put under regional theatre commands of the JNA that ignored the borders of republics and autonomous provinces after the 1988 reorganisation of the army in Yugoslavia. As a result, the presidency of republics lost the authority of commanding TO forces either in peace or in war. Spegelj 2001b: 89-93. In addition, as discussed in Chapter 4, the JNA forces seized the weapons stored for the TO forces in Croatia immediately after the elections in 1990, and thus the Croatian authorities could not use the arms of the TO in order to build up the national army. See Chapter 4 of this thesis (p. 95). 30 Jovan OpaCic, “Od sveca do izdajnika,”NIN, 1995/08/25: 25, cited in Ciger 2001: 202.

170 what would happen to the Serbs in Croatia in his first meeting with Milosevic in October 1990.31 Spegelj (2001a: 24), the then Minister of Defence of the new Croatian government, pointed out that the “Serb radicals had reckoned that they had absolute superiority, which was a delusion. They thought they had the

JNA under their thumb, and the TO forces of all the republic too via the general staff of the supreme command, while the police forces of the republics were technically inferior and smaller in number.” Hadzic (2004: 138), an expert on military affairs in Serbia, also argued that the strategy of the Serbian military-political elites “relied on the illusion of possessing enough power and force to impose solutions, i.e. to achieve their goals easily and quickly.”

This optimistic expectation turned out to be a serious miscalculation.

For example, despite the concentration of the best elite forces of the JNA and the tens of thousands of shells bombarded on the town, it took almost three months for the JNA forces to conquer Vukovar, which fell on 19 November

1991.32 While the Serb leaders expected that the military operations would be completed fairly quickly, the war in Croatia lasted more than six months and the Serbs could not achieve a decisive victory.

Two factors have been pointed out for the failure of the Serb rebels and the JNA forces to make a quick and decisive victory. Firstly, the morale among the JNA soldiers was low (Simac 2001: 31), and the response rate for the call up decreased rapidly. While the response rate to the call-up was very high (95-100%) in the first phase of mobilisation for the JNA from April to

June 1991, it dropped significantly to less than 60% in the third phase of mobilisation in September and October the same year (Marijan 2004: 296). The

31 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 13094 (2002/11/20). Babic claims that Milosevic repeated this phrase many times thereafter. See, for example, ICTY transcript of case IT-02-54 (Slobodan Milosevic): 13056 (2002/11 /19). 32 For die details of the fight in Vukovar, see Marijan 2004.

171 Belgrade opposition weekly reported in late September 1991 that only 50% of the reservists in Serbia and only 15% in Belgrade had obeyed orders to report for duty (Burg & Shoup 1999: 84). Kadijevic (1993: 97) himself admitted that

“mobilisation became the crucial limiting factor in achieving all plans.”

Secondly, the Croatian forces performed surprisingly well against the Serb rebels and the JNA forces. For example, the Croatian defence forces in

Vukovar exhibited a stiff resistance against the attacks of the JNA forces. At the end of September 1991, the Croatian forces occupied the JNA depots and garrisons and seized 250 tanks, 400-500 heavy artillery weapons, about

180,000 firearms and some 2 million tonnes of ammunition and other military hardware, which fundamentally altered the balance of military power (Spegelj

2001a: 34).

Even though it turned out to be a miscalculation, the Serb elite in

Croatia expected that the war would be over quickly once the JNA started its operations against Croatia with the superior firepower and manpower. In other words, they thought that the military resources they had at their disposal were superior to those of the Croatian government, and this belief significantly lowered the perceived cost of rebellion in the eyes of the organisers of the rebellion, while the actual cost of rebellion turned out to be much higher than expected. Due to the expected low cost of the rebellion, the Serb elite chose to take a violent option without trying harder to negotiate with the Croatian government and to achieve their objective by peaceful means.

4. Serbs in Bosnia

Finally, this section examines the effects of structural conditions in the case of Serbs in Bosnia. As for the motivating conditions, the actual

172 discrimination was not so manifest in the case of Bosnia due to the participation of the SDS in the coalition government, though some analysts point out the importance of fear of future discrimination as a motivator of rebellion. In the case of Bosnia, the key factor that motivated SDS politicians to take up arms was the moves for Bosnian independence. Here again, the dynamic grievances do not explain the occurrence of Serb rebellion well, since the Serb elites in Bosnia had chosen to take up arms sooner rather than later without trying to achieve their goal by peaceful means. By examining the effects of enabling conditions, this section will show that it was the newness of the Bosnian state that lowered the perceived cost of rebellion in the eyes of the

Serb leaders in Bosnia.

(1) Motivating Conditions Compared to the Serbs in Croatia, the level of discrimination in

Bosnia was much lower because the SDS-BiH participated in the coalition government after the elections in 1990 and thus there was no monopoly of power by one ethnic group after the elections. While the exclusion of ethnic minorities from the government occurred in some municipalities at the local level, the establishment of the three-party coalition government did prevent the exclusion of ethnic minorities at the national level. For example, in February

1991, the SDA and the HDZ-BiH proposed a declaration that stipulates that the republican laws precede the federal laws and confirms the right of secession of the republic from the federation, but this proposal was rejected due to the opposition of the SDS-BiH.33

While the actual discrimination was not so manifest as a motivating factor, some point out that the fear of future discrimination motivated the Serbs

33 RFE/RL Report on Eastern Europe, 2-27,1991/07/05: 30.

173 to take up arms. For example, Jovic (2001: 21-22) argues that “the fear of being a minority was, and remains, a major motivator of conflict in the former

Yugoslavia” and this fear “was created by nationalist members of the counter-elite.” Having recognised that the Serb leadership in Bosnia “led an effort which victimised countless innocent people,” Plavsic stated that they committed such terrible acts because of “a blinding fear” which led to “an obsession, especially for those of us for whom the Second World War was living memory, that Serbs would never again allow themselves to become victims. In the case of Bosnia, an important motivating factor was again the moves for independence. The critical moment in this regard was 15 October 1991, when the SDA and the HDZ-BiH proposed the “memorandum on the sovereignty of Bosnia.” The SDS-BiH members opposed it, but the SDA and the HDZ-BiH ignored their opposition and adopted it according to the majority principle after the SDS-BiH members withdrew themselves from parliament.35

Since then, the SDS-BiH rapidly made a series of moves for the division of

Bosnia, such as the establishment of the “Serb National Parliament of Bosnia” in late October, the organisation of the referendum among the Serb population in Bosnia to ask whether they wish to remain in Yugoslavia in November, and the establishment of the “Serb Republic of Bosnia and Herzegovina” in

January 1992.36 When the Bosnian government conducted a referendum on the independence of Bosnia at the beginning of 1992, the Serbs boycotted the referendum. The moves for Bosnian independence, therefore, were a critical motivator for the Serb leaders to take up arms.

34 ICTY transcript of case IT-00-39 & 40/1 (Biljana PlavSic): 609-610 (2002/12/17). 35 As for the original text of the memorandum as well as the counter proposals by die opposition parties, see Borba 1991/10/16. 3 Burg & Shoup 1999: 97. For the details of the development after the memorandum, see, e.g., Burg & Shoup 1999: 69-120.

174 (2) Enabling Conditions

Let us now examine the effects of enabling conditions in Bosnia. In the case of Serbs in Bosnia, many of the enabling conditions do seem explain the decisions by the SDS-BiH leaders to take up arms. The level of economic development was low: while the data of GDP per capita is not available for the

1990s, Bosnia was a poorer republic in the socialist Yugoslavia, ranking third poorest after Kosovo and Macedonia in 1988 (PleStina 1992: 180-181), and was experiencing serious economic hardship before the civil war started in

1992. Its geographic terrain is highly mountainous: 60.5% of its territory is coded as mountainous, which is much higher than the mean value of the Fearon and Laitin dataset. The Serbs were concentrated in some areas and they also had rural bases in Bosnia. Bosnia experienced a transition from the one-party regime to the multi-party regime in 1990, even though this transition is not captured by the polity2 dataset since Bosnia enters Polity IV dataset only in 1992 (if one assume the same value for all republics in Yugoslavia, the value of the polity2 variable for Bosnia (Yugoslavia) is -5 since 1980 to 1991 and it is 0 for Bosnia in 1992, thus the variable of political instability would take a value of 1).

As in the case of Croatia, one important enabling condition was the newness of the Bosnian state. Just like the Croatian government, Bosnia did not have its own army and had to build an army from scratch. On the other hand, the Serbs had a large and professional army (the JNA) on their side, as well as their own police force and a number of paramilitary forces from Serbia. This perceived weakness of the Bosnian state, combined with the perceived superiority of the military resources of the Serbs, affected the decisions by the

SDS-BiH politicians to take up arms earlier rather than later. For example,

175 Karadzic stated as follows already in October 1991, implying his confidence in the superiority of Serbs against the Bosnian government led by Muslims: “Do not think that Bosnia would not fall into the hell. Muslims may disappear altogether, because they cannot defend themselves in war.”37 In November

1991, when he made a speech to the Serb mayors in Bosnia during the preparation for the referendum, Karadzid said: “Now we have an army, which is three times stronger ... This army has weapons, technology, other resources, and huge war reserves. It has commanding and leading cadres and has its goal, which is identical to your goal” (Ibrahimagid 2001: 250). In a private conversation with Gojko Djogo, the then president of the Association of the

Serbs in Belgrade, Karadzic said “they [Muslims] must know that there are 20,000 armed Serbs around Sarajevo...they will disappear! Sarajevo will be a melting pot in which 300,000 Muslims will die.”38 Nikola Koljevic, a Serb member of Bosnian presidency, was reported to have suggested that the whole military operation would be over within ten days (Burg & Shoup 1999: 130).

The SDS-BiH politicians believed that they were superior to the Bosnian government with regard to military resources, and this belief affected the decisions to take up arms earlier without trying harder to achieve their goal by peaceful means.

In addition, the hard nature of the international borders also encouraged the Serb elite to take up arms earlier. The Serb elite in Bosnia felt that they had to take an action quickly to challenge the unity of Bosnia because, if they did not do so, the international border of Bosnia would be recognised and consolidated, and it would be difficult for them to challenge and change the borders later. For example, in the speech made in November 1991

37 NIN, 2002/08/29: 16. 38 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 13074(2002/11/20).

176 mentioned above, Karadzic said: “if we wake up one day... in Alija

[Izetbegovic]’s state, with borders recognised by Europe, we should know that such borders could never be changed any more after such recognition”

(Ibrahimagic 2001: 244). In other words, the Bosnian Serbs took up arms immediately after the declaration of independence by the Bosnian government not only due to the perceived weakness of the military of the new state (as argued by Fearon and Laitin, for example) but also due to strategic calculations on the international recognition of borders: the later the rebellion is, the more difficult to challenge and change the borders of the state.

Conclusion

This chapter examined why the rebels decided to take up arms in the four ex-Yugoslav cases. The present chapter systematically examined motivating factors and enabling conditions, as well as the dynamic aspect of grievances, based on the theoretical arguments and empirical findings of the previous chapter. As for the motivating factors, the presence of discrimination was manifest in three out of four cases (Kosovo, Macedonia and Croatia), and the loss of autonomy or the issue of stateness was important in three out of four cases (Kosovo, Croatia and Bosnia). As for the enabling conditions, the presence of harsh authoritarian regime was important in one case (Kosovo until

1996) and the newness of the state was significant in two cases (Croatia and

Bosnia). While not all countries examined here can be classified as

“mountainous” country, the area where rebellions occurred was relatively mountainous in all four countries (Kosovo in Serbia, Western part of

Macedonia, Knin area in Croatia, and Bosnia). As for the level of economic development, three countries (Serbia, Macedonia and Bosnia) were not so

177 highly developed. In addition, while Croatia was relatively more developed, the Knin area was a less developed region in Croatia. In all cases, ethnic groups were regionally concentrated and had some rural base in the country.

It is not the purpose of this chapter to single out one of these factors as a decisive explanatory factor. As discussed in Chapter 5, the present thesis argues for the complementary nature of the explanatory factors. The qualitative case study in this chapter supports this argument. Both motivating and enabling conditions were found important in these four cases, and this finding suggests that these two types of explanation are not mutually exclusive. The combination of factors that were important for the explanation of the occurrence of rebellion varies across the four cases, which implies that there is no singular combination of factors that is applicable to all cases of ethnic rebellions.

As for the dynamic aspect of grievances, this chapter has shown that it was an important factor for explaining the timing of the rebellions in Kosovo and Macedonia, where the structural conditions cannot fully explain the timing of the rebellion. This factor, however, was not so important in the case of Serbs in Croatia and Bosnia, where the Serb leaders chose to take up arms earlier rather than later. As discussed in Chapter 5, this difference is due to the type of decision-makers. If the organisers of the rebellion have a large amount of resources at their disposal, it could lower the perceived cost of the rebellion, and as a result, they might choose to rebel before trying to achieve their goal by peaceful means. In the case of Serbs in Croatia and Bosnia, they had a large amount of military resources at their disposal while the Croatian and Bosnian state authorities were perceived as weak and inferior to the Serbs, and this perception led the SDS politicians to take up arms earlier. This observation supports the findings of the large-N analysis in Chapter 5 that the dynamic

178 aspect of grievances does not have a significant effect on the occurrence of large-scale rebellion, while it does have a significant effect on the occurrence of small-scale rebellion.

The case study of this chapter pointed out one important factor that was not included in the theoretical framework of Chapter 5, namely the availability of arms. In the case of Albanian rebellions in Kosovo and

Macedonia, this was one of the crucial factors which explain the timing of the occurrence of rebellion. In the case of Serbs in Croatia and Bosnia, this factor also explains the early decision to take up arms, as discussed above. This factor was not included in the discussion of Chapter 5 because it is extremely difficult to operationalise and quantitatively measure in a convincing manner. To test the validity of the argument for the importance of the availability of arms, therefore, remains a task for the future research.

179 Part IV: Dynamics of Rebel Escalations

Chapter 7 State Repression as an Intervening Variable: Quantitative Analysis of Rebel Escalations

Introduction

This chapter will examine the dynamics of escalation. As argued in

Chapter 1, explaining the dynamics of escalation is critically important for the theory of large-scale rebellion if the rebellion starts from a low intensity. Because of the unitary-actor assumption and the binary coding of the occurrence of large-scale conflict, the issue of escalation has not been explored fully and systematically in the existing literature. By theorising and empirically examining the dynamics of escalation, therefore, this chapter attempts to make some contributions to the study of ethnic conflict.

This chapter consists of two parts. The first part presents the theoretical argument about the dynamic relations between the occurrence of initial low-intensity rebellion, state reactions and the escalation of rebellion. It will then discuss some structural factors that are to be included in the statistical analyses as control variables. The second part conducts large-N analyses to test the hypotheses drawn from the theoretical argument. It will firstly discuss research designs, including operationalisation and estimation methods. It will finally present and discuss the empirical results of the large-N analyses.

180 1. Theoretical Exploration

(1) Dynamics between the Ethnic Group and the State

Why do some ethnic groups experience an escalation of rebellion after the onset of low-intensity rebellion, while others do not? One of the answers to this question lies in dynamics between the ethnic group and the state authorities.

In many cases, the occurrence of low-intensity rebellion in the country is perceived by the state authorities as a signal that the security and integrity of the state is being threatened. Facing such a situation within the state, state institutions such as the government, police and military would consider taking some measures in order to eliminate the threat to the country. The nature of this reaction made by the state authorities against the initial onset of rebellion is of critical importance for the escalation or de-escalation of the conflict.

The option often considered by the state authorities facing such rebellion is repression. The state authorities try to eliminate the problem by suppressing the rebel activities by force or even by physically destroying the rebel organisations. If successful, this option will “solve” the problem without making any concession to the rebels, however superficial this “solution” is.

When repression by the state authorities is excessive, however, it can be counterproductive and radicalise the ethnic group that is targeted by the state. Indeed, many authors have pointed to the counterproductive effect of state repression. For example, analysing the sources of uprising by Kashmiri people against the Indian State, Bose (2003:116) points out that the “regime of repression had the effects of further radicalising public opinion and of convincing thousands of Kashmiri youths to take up arms to fight the Indian state.” In the case of the Kurdish question in Turkey, it was pointed out that the security operations and the practice of village burning was fuelling Kurdish

181 nationalism and was forcing especially young people to join the ranks of the rebels (Kiri§ci & Winrow 1997: 131). In Sri Lanka, the police and military responded brutally when the LTTE started its low-intensity rebellion, and the military terror and repression directed against the Tamil population played a

“vital catalytic role” and contributed to the ascendancy of the LTTE.1 The counterproductive nature of the coercion is pointed out almost 30 years ago by

Hibbs (1973: 182), who concluded that “the nearly instantaneous response to repression is most often more mass violence” based on the results of his large-N analyses.

Why do repressions radicalise the ethnic group rather than lead to decisions to give up arms? Firstly, if the repressive policies lead to serious casualties or physical damage of properties, they will increase grievances among members of the ethnic group. If one loses one’s close friends or family members, one can easily be radicalised and may decide to take up arms. They can also be radicalised by the fact that the state has killed their “fellow” people of the same ethnic group. While these people did not take up arms when the initial rebellion occurred, now they may join the rebels and take up arms due to the consequences of repressive measures. Secondly, the repressive measures would lead to the loss of confidence in the government as their “representative” government among the members of the ethnic group. They would start thinking that the state authorities would not care and protect them but rather would harm them. The loss of confidence in the state authorities and the government eventually will lead to the increase in the support for their “own” statehood or the establishment of their self-rule.

1 Bose 1994: 91. Bose called this dynamics as “the dialectic of state repression and nationalist resistance.” See, Bose 1994: 89. For the further detail of this dialectic, see Bose 1994: 92-116. The figure of the LTTE membership has grown from mere 30-odd individuals in July 1983 to some 4,000 in July 1987 (Bose 1994: 87).

182 If the repressive measures taken by the state authorities led to the radicalisation of more members of the ethnic group and thus led to the escalation of the ethnic conflict, here is a dynamic relationship between the onset of lower-intensity conflict, the repressive measures taken by the state authorities and the escalation of rebellion: the state authorities take repressive measures because of the onset of lower-intensity conflict, and anti-state rebellion escalates because of the repressive measures. In other words, the state repression is the key intervening factor which explains the occurrence (or non-occurrence) of the escalation of rebellion. This relationship is illustrated in

Figure 7.1.

Figure 7.1 Dynamic Relationship between Rebellion and Repression

Repressive State Authorities policies

Ethnic Lower-intensity Large-scale Group Rebellion Rebellion

This dynamic relationship is the main hypothesis tested in this chapter.

In order to test it, it is necessary to establish two links: firstly, the onset of lower-intensity rebellion leads to more repressive measures of the state authorities, and secondly, the repressive measures taken by the state authorities lead to the escalation of rebellion. Therefore, one can draw the following

183 hypotheses:

Hypothesis 1: If a low-intensity rebellion occurs in a country, the state

authorities are more likely to take more repressive measures against

rebels.

Hypothesis 2: As the level of state repression becomes higher, the level of

intensity of rebellion in a country is more likely to increase.

Of course, this argument does not mean that these dynamic relations between ethnic groups and state authorities are the only determinants of the escalation of rebellion. One can consider some structural factors that are likely to affect the repressive policies of the state authorities and the escalation of rebellion. These structural factors will be treated as control variables in the large-N analyses below. Let us now examine these structural factors.

(2) Structural Factors for State Repression

This section examines some structural factors that might affect state repression. Firstly, one can argue that the political regime of the country may affect the level of state repression. For example, analysing the determinants of the level of political repression (such as suspension of civil rights and liberties),

King (2000) found that democracy has “pacifying effects” on the behaviour of repressors towards dissidents: while the increase of political dissent tends to increase the level of political repression when democracy is weak, it does not increase the level of political repression so much when democracy is strong. A similar argument can be made for the military repression. In other words, the level of state repression may be lower in a country under a democratic regime, because it would be more difficult to resort to the arbitrary use of

184 police/military force against its population if there are strong democratic constraints on the actions of the government. Thus, one would expect that it is less likely that the state authorities take more repressive measures against the rebels as the political regime gets more democratic in a country.

Secondly, one can argue that GDP per capita might affect government decisions on military repression. If a country enjoys a higher GDP per capita, the government would try to avoid the risk of disturbing the normal economic activities and of earning a bad reputation among the international community by starting military operations against its population within the country. If a country is poor, on the other hand, there will be much less at risk compared to rich countries, and in addition, poorer countries lack sufficient resources to share with rebelling groups to resolve the conflict more peacefully. Therefore, one may expect that the government in poorer countries is more likely to take repressive military measures compared to richer countries.

These two factors will be used as control variables in the large-N analyses of the level of state repression below.

(3) Structural Factors for the Escalation of Rebellion

Let us now examine some structural factors that might affect the escalation of rebellion. Recall that motivational factors and enabling factors were used as structural conditions that affect the decisions of those who take up arms against the state. As for these structural conditions, one can argue that they also might affect the decisions of those who take up arms later. For example, if the presence of discriminatory policies motivated those who took up arms at the beginning, they may well motivate those who take up arms later, causing an escalation of the rebellion. If the weakening of the centre enables the organisers of rebellion to take up arms, it will also enable more people to

185 take up arms for the same reason, leading to an escalation of rebellion. If the structural factors that affected the occurrence of initial rebellion work under the same logic, they will have a similar effect for the escalation of rebellion.

However, the automatic application to the escalation of rebellion does not sound equally plausible for some factors. Firstly, the variable of the dynamic aspect of grievances, operationalised by the duration of the protest activities, may not affect the occurrence of escalation, because the peaceful protest activities might cease to continue once the rebellion is started by those who take up arms against the state. If protest activities cease to continue after the occurrence of initial rebellion, the duration of protest activities will be zero when the escalation occurs and thus will have no impact on the occurrence of escalation. Secondly, variables of new states and political instability at the centre may not have a significant effect on the occurrence of escalation. By definition, it can only explain the events that happen within a short period: the newness of the state variable can only explain the events that happen within two years from independence, and the political transition variable can only explain the events that occur within three years since the change in the political regime indicator takes place. If the escalation takes more than 3-4 years since the onset of rebellion, therefore, these variables will naturally take the value of zero and will have no impact on the occurrence of escalation.

One should also note that the ethnic groups that have never experienced an occurrence of rebellion during the period from 1945 to 2000 will be excluded from the analysis below, because it attempts to explain the variance between the occurrence and non-occurrence of escalation among the ethnic groups that experienced the occurrence of initial rebellion. Therefore, the mean of the explanatory variables may be skewed among the ethnic groups that have experienced an occurrence of initial rebellion, and the variance of

186 explanatory variables may be much more limited compared to the overall sample. It is not surprising, therefore, if the structural factors used in Chapter 5 do not have a significant effect on the escalation of rebellion.

Despite these reservations, the structural factors used in Chapter 5 will be used in the analysis below. The purpose of the inclusion is twofold. Firstly, these structural conditions can be used as control variables in the analysis to test the explanatory power of state repression. If state repression has a significant effect on escalation even when these structural conditions are controlled, one can be more confident in the results of the large-N analysis.

Secondly, one can check whether these structural conditions have a significant effect on the escalation as well. In other words, by including these variables, one can see whether it is plausible to assume that the structural factors that affect those who take up arms at the beginning will have a similar effect on the decision made by the people who take up arms later.

2. Large-N Analysis

This section conducts large-N analyses to test the hypotheses presented above. It will firstly discuss research designs, including operationalisation and estimation methods, and then present and discuss the empirical results of the large-N analyses.

(1) Research Design

Operationalisation of the State Repression

In order to conduct a statistical analysis, there must be some variable that indicates the seriousness of the repressive measures taken by the state. In the present thesis, some of the “government repression of the group” variables

187 in the MAR dataset are used as a proxy for the repressive measures by the state authorities. Among 23 variables of “government repression” in the MAR dataset phase El, three are directly related to the military operations of the government forces against the rebels: REP20 (Military campaigns against armed rebels), REP21 (Military targets and destroys rebel areas) and REP22

(Military massacres of suspected rebel supporters). Each variable is coded on the 4-point scale, from 0 to 3 (as for the scale on which this variable is coded, see Appendix 2), and the larger value of these variables indicates that the harsher and more indiscriminate repressive military operations are conducted against the ethnic group by the government forces.

The “repression” by the state authorities covers a various range of activities by the government, police and military, and the “government repression” variables in the MAR dataset also include other kind of policies such as political arrests or confiscation of properties. However, the military aspect of repressive policies is used here because it often radicalises members of the ethnic group most dramatically. The theoretical argument in this chapter assumes that a larger number of members of the ethnic group will be radicalised and join the rebels when they are physically attacked by the government forces and lose fellows of the same group, their own friends and families. The variables of the military repression, therefore, are most suitable to test the main hypotheses in this chapter.

One disadvantage of using these variables is that they are included in the dataset only in the Phase in of the MAR project, and thus data exists only for the period of 1996-2000. Therefore, any results of the statistical analysis using these variables are confined to this period and we cannot make any inference on the entire period after the WWII. The results of the statistical analysis below are thus preliminary at best. However, it is still worth

188 conducting such a statistical analysis for several reasons. Firstly, even when we deal with the five-year panel data (the data is coded annually), the dataset still contains a sufficiently large N (more than 600) for the statistical analysis to be meaningful. Therefore, as far as the size of the dataset is concerned, there is no technical reason to stop conducting a statistical analysis. Secondly, the coding of the “seriousness” of the state repression against all ethnic groups in the dataset (more than 300 groups in 125 countries) requires an enormous amount of work, and the variables included in the MAR dataset are significant resources for researchers. Therefore, conducting a statistical analysis with the available dataset is at least better than throwing out the baby with the bathing water.

In order to test the hypotheses set above, two separate statistical analyses will be conducted. In both cases, the unit of analysis is again ethnic group-year, and the dataset covers all ethnic groups in the MAR dataset and the period from 1996 to 2000 (i.e. 5-year panel). Research designs of these analyses are discussed below.

Effect of Lower-intensity Rebellion on State Repression

Firstly, a statistical analysis will be conducted to test hypothesis 1 above. In other words, it analyses whether the occurrence of lower-intensity rebellion incites the military repression by the state authorities. Therefore, the dependent variable here is an index of military repression (REP20, REP21, or

REP22) against the ethnic group in a particular year y. Note, however, that the value of the military repression index in a particular year y may simply be a continuation of the level of repression in the previous year (y-1). In order to distinguish the mere continuation of the repression from the increase of the level of repression, the lagged dependent variable (LDV), i.e. the level of

189 repression in the previous year, will be included in the analysis as a control variable.

The independent variable of theoretical interest here is the occurrence of lower-intensity rebellion of the previous year (y-1). The result expected by the theoretical argument above is that the occurrence of lower-intensity rebellion of the previous year has a statistically significant effect of increasing the level of military repression of the subsequent year (i.e. the coefficient is positive).

In order to capture the occurrence of lower-intensity rebellion, two variables are constructed based on the anti-regime rebellion index. The first variable takes the value of the rebellion index of the previous year if it is equal to or smaller than 5. It will be coded as zero if the level of rebellion in the previous year is 6 or 7 (LOWREB hereafter). This variable thus captures not only the occurrence of lower-intensity rebellion but also the level of the rebellion in the previous year. The second variable is a dummy variable that takes the value of 0 when the rebellion of the previous year is zero, and takes the value of 1 when the rebellion of the previous year is from 1 to 5

(LREBDUM hereafter). If the level of rebellion in the previous year is 6 and 7, it will also be coded as zero. In other words, this dummy variable captures the occurrence of lower-intensity rebellion in the previous year, excluding the occurrence of the higher-intensity conflict. These two independent variables are used in order to see the effects of the lower-intensity rebellion in the previous year on the level of state repression.

One problem for both of these operationalisations is that instances of non-occurrence of rebellion (REB = 0) and occurrence of high-level rebellion

(REB = 6, 7) are confused and cannot be distinguished from each other. In order to solve this problem, a dummy variable is constructed that takes the

190 value of 1 if the level of rebellion in the previous year is either 6 or 7 and included in the analysis as a control variable (HIGHREB hereafter). By including this dummy variable, one can evaluate the effects of the instances of low-intensity rebellion in a more appropriate manner, without confusing instances of non-occurrence of rebellion and large-scale rebellion.

As for the control variables, two variables that were used in Chapter 3 will be used. The first is POLITY2, an index of political regime of the country in which the ethnic group live, taken from the Polity IV dataset. The second is the GDP per capita of the country in which the ethnic group live. Based on the arguments above, one would expect that the variable POLITY2 has a statistically significant effect and its coefficient will have a negative sign (the more democratic the regime is, the less likely it is that the level of state repression becomes higher). As for the GDP per capita, one would expect that the variable GDP has a statistically significant effect and its coefficient will have a negative sign (the higher the GDP is, the less likely it is that the level of state repression becomes higher).

As for the estimation method, the ordered logit model is used again.

As discussed in Chapter 3, this estimation method is the most appropriate when the response variable is ordinal and categorical (see Chapter 3). The fitted model of the present analysis can be summarised as follows:

1 1 Pr (REP20/ = m | x) l1 + i expv(-Tm-1+ ^ X b) ' where

Pr (REP20/ = m ) = probability that the level of military repression

against the ethnic group i takes the value of m (m = 0 to 3) xfi = (fij LDVi + p2 LOWREB/ + p3 HIGHREB/ + p3 POLITY2/ + p4

GDP/).

In the formula above, REP20 may be replaced by REP21 or REP22 and LOWREB may be replaced by LREBDUM to use the alternative operationalisations, The statistical significance of the coefficients will be calculated based on the robust standard error, which takes into account the panel structure of the dataset by clustering the observations of the same ethnic group. By doing so, one can avoid obtaining an invalid standard error due to the invalid assumption that all observations are mutually independent.

Effect of the State Repression on the Escalation of Rebellion The second analysis will test hypothesis 2 above. This analysis will be conducted in order to see whether the higher level of military repression leads to the escalation of rebellion. In order to operationalise the escalation of rebellion, two methods are used in the present analysis.

Firstly, a dummy variable is constructed that takes the value of 1 whenever the value of the rebellion index increases from the previous year, and takes the value of 0 otherwise (ESC hereafter). This variable thus captures the instances of escalation. The advantage of this operationalisation is that it is possible to pinpoint the moments of escalation. The disadvantage of this operationalisation is, however, that it cannot take into account the degree of escalation. For example, the increase of the rebellion index from 1 to 2

(increase by 1) and that from 1 to 7 (increase by 6) would be treated equally as an instance of “escalation.” In other words, one cannot distinguish the relatively modest escalation from more serious one.

Secondly, therefore, the value of the rebellion index will be used as a

192 dependent variable (REB hereafter). One problem with using the value of the rebellion index as a dependent variable, however, is that one cannot distinguish the instances of escalation from instances of mere continuation of rebellion. As a remedy to this problem, therefore, the lagged dependent variable (LDV, i.e. the value of the rebellion index of the previous year) will be used in the analysis as a control variable. By including the LDV in the model, one can estimate the effects of explanatory variables on the escalation of rebellion more properly.

Note that one cannot use the entire data in the MAR dataset here. If one conducts an analysis on the “escalation” using the entire data in the dataset, the occurrence of initial rebellion will be also included in the dependent variable. Since one is interested in the escalation of rebellion that occurs after the onset of rebellion, this first “onset” must be excluded from the dependent variable. In order to limit the scope of analysis to the observations after the first “onset” of rebellion, all observations before the occurrence of initial rebellion and instances of initial rebellion are omitted from the dataset This naturally leads to the omission of all ethnic groups that remained peaceful throughout the period from 1945 to 2000.

The independent variable of key interest here is the level of military repression in the same year. The three variables discussed above, REP20,

REP21 and Rep 22 will be used as independent variables. In addition, the sum of these three variables (REP20 + REP21 + REP22) is also used as an independent variable (REPTOTAL hereafter) in order to see the effect of the overall level of repression on the level of rebellion. If the argument in this chapter is right, the higher level of military repression will have an effect of increasing the level of the rebellion. The expected sign of the coefficient of the military repression is, therefore, positive.

193 As discussed above, the variables used in Chapter 5 will be used as control variables. The NEWST variable (newness of the state), however, had to be dropped from the analysis because of the lack of variance: during the period from 1996 to 2000, no country in the dataset became independent and it takes the value of zero for all countries and years. All other variables, namely

PROTDUR (duration of the protest activities), AUTLOST (loss of autonomy),

POLDIS (political discrimination), ECDIS (economic discrimination), DICT

(dictatorship), MONT (mountainous terrain), INSTAB (political instability at the centre), GDP (GDP per capita) and REGCON (regional concentration of the ethnic group) are included in the large-N analyses conducted below.

As for the estimation method, different types of methods are used for two dependent variables. For the escalation dummy variable, the random-effects logit model is used, because of the binary coding of the variable and the panel structure of the dataset (see Chapter 5 for more details on this method). For the rebellion index variable, the ordered logit model is used because of the ordinal and categorical nature of the variable. Here again, the statistical significance of the coefficients will be calculated based on the robust standard error, which takes into account the panel structure of the dataset by clustering the observations of the same ethnic group. The fitted models can be summarised as follows.

For the escalation dummy variable: logit y — fio + U, + PiREPlQij + /^PROTDURy + /%POLDIS,y + /^ECDISy +

AAUTLOSTy + fisDlCTij + /?7INSTAB,y + jfcGDP*- + fiMOKTy +

/foREGCONy

194 where

logit jj = log odds that the ethnic group /' experiences the escalation of

rebellion in the year j

the random effect component { U, } are independent N (0, a 2) variates.

For the rebellion index variable:

1 1 Pr (REB/ = m | x)

where

Pr (REB/ = m) = probability that the level of rebellion by the ethnic group

/ takes the value of m (m = 1 to 7)

xj3 = (fijLDVi + #>REP20z + /TjPROTDURz + /^POLDISz + /?5ECDISz +

AAUTLOST/ + j37DlCTi + /&INSTAB/ + frGDPi + /?/0MONT/ +

AyREGCON/).

(2) Empirical Results

Effect of Lower-intensity Rebellion on the State Repression

The results of the analysis on the effect of lower-intensity rebellion are shown in Table 7.1. In Model 1 and 2, REP20 is used as a dependent variable.

Model 1 is a model with LOWREB as an explanatoiy variable. The results show that the level of lower-intensity rebellion in the previous year has a statistically significant effect on the level of military repression, and its coefficient is positive as expected by the theoretical argument. Model 2 is a model with LREBDUM as an explanatory variable. The occurrence of lower-intensity rebellion in the previous year again has a statistically significant effect on the level of military repression, and its coefficient is

195 positive. These results thus support hypothesis 1. In both models, the level of

repression in the previous year (LDV) has a statistically significant effect and

its coefficient is positive. This means that the level o f repression tends to

persist: the state authorities that took repressive measures in the previous year

are more likely to take similar measures in the next year. In both models, the

occurrence of high-intensity rebellion in the previous year also has a

statistically significant effect, and its coefficient is positive. This means that the

state authorities are more likely to take repressive measures when a large-scale

rebellion has occurred in the country (which is, o f course, understandable).

Table 7.1 Ordered Logit Analysis of the Determinants of Military Repression, 1996-2000 (Dependent Variable: REP20) mod el 1 mod el 2 Independent Variables coef std err coef std err Repression of the Previous Year 4.80*** 0 43 4.62*** 044 (LDV) Rebellion of the Previous Year 0 28*** 0 1 Low-intensity Rebellion in the 1.34*** 0.37 Previous Year High-intensity Rebellion in the 1.71*** 0 48 2.00*** 0 46 Previous Year Democracy 0.0057 0.02 0.015 0.021 o oo * o o Level of Economic Development 1 0045 -0.085* 0.045 Cut Point (1) 3.62 0 3 3.8 0.31 Cut Point (2) 9 65 0.87 9 66 0.81 Cut Point (3) 13.44 1.55 13.33 1.5 N 968 968 *p- 0.1. ** p <0.05; ***p<0 01

As for the two structural variables, the effect o f the political regime is not statistically significant in both models, and thus there is no empirical evidence for the argument on the effect of democracy. The effect of the level of economic development (GDP per capita), however, does have a statistically significant effect on the level of state repression at 0.1 level in both models,

196 and the sign of the coefficient is negative as expected by the theoretical argument. This means that the state authorities in a more developed country are less likely to take repressive military measures against the rebels.

While the results of the analysis using REP21 as a dependent variable do not support the hypotheses of this chapter, those using REP22 as a a dependent variable also support hypothesis 1. One can conclude, therefore, that there is some empirical evidence that the occurrence of lower-intensity rebellion affects the level of military repression taken by the state authorities.

Let us now examine the effect of the state repression on the escalation of rebellion.

Effect of the State Repression on the Escalation of Rebellion The results of the random-effects logit analysis of the occurrence of escalation (ESC) are presented in Table 7.2. Model 1 is a model using REP20 as an independent variable. REP20 has a statistically significant effect on the escalation of rebellion, and the sign of the coefficient is positive, as expected.

In other words, the harsher the military repression is, the more likely that the rebellion escalates into the larger-scale one, other conditions being equal.

While REP21 and REP22 do not have a similar effect on the escalation of rebellion,3 the overall level of military repression (REPTOTAL) also has a statistically significant effect on the escalation of rebellion, and its coefficient is again positive (see Model 2 in Table 7.2). These results support hypothesis 2 of the present chapter. As for the control variables (structural conditions), most of the variables remain statistically insignificant. Only exception here is the

The results of the analyses using REP21 and REP22 are not shown here due to the lack of space. See Appendix 3 for the results. 3 The results of the analyses using REP21 or REP22 as an independent variable are not shown here due to the lack of space. See Appendix 4 for these results.

197 level of economic development (GDP), which consistently has a statistically significant effect on the escalation of rebellion. Its coefficient is negative, as expected by the theoretical argument.

Table 7.2 Random-Effects Logit Analysis of Determinants of the Escalation of Rebellion. 1996-2000 Model 1 Model 2 coef std .err. coef std en Military Repression (REP20) 0.87*** 023 Total Level of Repression 020* on (REPTOTAL) Political Discrimination -0019 0 14 0 026 0 13 Economic Discrimination 021 0.15 0.22 015 j j i ■ 0 0 Lost Autonomy 1 0 16 1 0 16 Duration ofPeacefi.il Protest -0 009 001 -0 007 0.01 Dictatorslup -0 16 0 34 -0 17 0 34 Mountainous Terrain -0 043 0 11 -0024 O il Political Instability 0.19 0 35 0.15 0 34 Level of Economic Development -0 15** 0 06 -018*** 0 06 Regional Concentration -0 49 0 44 -0 48 044 Constant -1.57** 0.6 -1 40** 059 N 660 659 *p 0 1. ** p < 0 05; ***p<0.01

The results of the ordered logit analysis using REB as a dependent variable are presented in Table 7.3. Model 1 is a model using REP20 as an independent variable, and it has a statistically highly significant effect on the level of rebellion, even when the level of rebellion in the previous year is controlled for. REP21 also has a statistically significant effect on the level of rebellion, while REP22 does not.4 Furthermore, here again, the overall level of military repression (REPTOTAL) has a statistically significant effect on the level o f rebellion, and its coefficient is positive (see Model 2). These results again support hypothesis 2. As for the structural conditions, most of them

4 The results of the analyses using REP21 or REP22 as an independent variable are not shown here due to the lack of space. See Appendix 5 for these results.

198 remain statistically insignificant in both models, including the level of economic development. The only exception here is the duration of peaceful protest activities (dynamic grievances): in all models, it consistently has a statistically significant effect on the level of rebellion, and its sign is positive as expected.

Table 7.3 Ordered Logit Analysis of the Determinants of Level of Rebellion, 1996-2000 (Dependent Variable: REB) Model 1 Model 2 coef. st d err coef std.en. Level of Rebellion ui the Previous 1.00*** 0 11 1.03*** 0.12 Year (LDY) Military Repression (REP20) 1 48*** 0 27 Total Level of Military Repression 0 59*** 0.17 Political Discrimination 0.06 0 11 0.07 011 Economic Discrimination -0 02 0 11 0 0 12 Lost Autonomy -0.07 0 12 -0 07 0 11 Duration of Peaceful Protest 0 012** 0 005 0 012** 0006 j

Dictatorship P i 0 26 0 0 29 Mountainous Terrain -0 08 0 1 -0 06 0 08 Political Instability 0 2 0 24 0 12 0 23 Level of Economic Development -0 01 0.03 -0 02 0.03 Regional Concentration 0 22 0.34 0 12 0 39 Cut Point (1) 2 14 045 2 18 0 47 Cut Point (2) 3 23 046 3 21 0 48 Cut Point (3) 3 88 0.51 3 83 0.53 Cut Point (4) 5 01 0 56 49 0.58 Cut Point (5) 6.38 0.65 6 23 0.65 j ■ CO Cut Point (6) 7 38 0.72 i 0.75 Cut Point (7) 857 083 86 087 N 659 658 *p<01. **p <■ 005: ***p<0.01

In general, most of the structural conditions remain statistically insignificant on the occurrence of escalation or the level of rebellion after the initial onset. This means that one should not assume that the factors that explain the onset of initial rebellion equally explain the escalation of rebellion.

On the other hand, the variables of military repressions against the rebels have a statistically significant effect on the level of rebellion more consistently

199 across different models. The results of these analyses thus suggest that there is indeed a dynamic relationship between low-intensity rebellion, state repressions and larger-scale rebellion. The occurrence of low-intensity rebellion is likely to incite military repressions by the state authorities. These military repressions, however, tend to be counterproductive: the higher the level of military repressions is, the more likely that the anti-state rebellions escalate into the larger one, at least in the short-term.

One can see this counter-productive nature of the military repressions vividly when one makes a two-way tabulation of the rebellion index and the total level of military repressions against the ethnic group (REPTOTAL), which is shown in Table 7.4. The anti-state rebellion tends to be more intense and larger-scale as the level of military repressions becomes higher. While members of the ethnic group may remain peaceful under relatively modest repression (17 groups remain peaceful under the value of 1 of the military repression, and one group remain completely peaceful under the value of 2 of the military repression), no ethnic group remains peaceful against harsher and more indiscriminate military repressions. Under the highest level of military repressions observed in the period of 1996-2000, all ethnic groups were involved in the large-scale anti-state rebellion (6 or 7 of the rebellion index).

200 Table 7.4: Cross-Tabulation of Rebellion Index and the Level of Repression (Number of Instances, 1996-2000) Total Level of Military Repression (REPTOTAL) 0 1 2 3 4 5 6 7 Total 0 411 17 1 0 0 0 0 0 429 1 53 21 0 -> 0 0 0 0 76 o 20 12 1 1 0 0 0 0 34 Rebellion 3 33 11 3 n -> 0 0 0 51 index 4 31 4 1 *7 5 1 0 56 5 0 23 0 3 0 0 0 28 6 16 -> 4 0 3 *■) 31 n 8 13 5 *> 6 3 *) 6 45 Total 534 144 16 12 r)m) 8 6 8 750

It is obvious that state repression is not a necessary condition of the escalation of the ethnic conflict, since some ethnic groups were involved in large-scale rebellion even when the military repressions are virtually absent.

Table 7.4 implies, however, that harsh and indiscriminate state repressions might be a sufficient condition for the escalation of ethnic rebellion: whenever the state authorities take harsh and indiscriminate military repressive measures against an ethnic rebellion, it is highly likely to induce an even larger anti-state rebellion by the members o f the targeted ethnic group.

In a word, the dynamic relationship between the ethnic group and the state is a spiral of violence, where the degree o f violence of the ethnic rebellion is intensified by the degree o f violence o f the state repressions and vice versa.

However, this does not mean that the escalation of violence is an inevitable and automatic consequence of the occurrence of lower-intensity rebellion. The fact that repression is more likely to be harsher when the lower-intensity rebellion occurs does not mean that all state authorities will take repressive measures.

Which is chicken and which is egg, if the argument above holds? One may argue that, if there is a dynamic relationship, it would be difficult to figure

201 out which occurs first and which follows. It seems, however, that the onset of lower-intensity rebellion tends to precede the military repressions. As far as the

MAR dataset is concerned, the military repressions are extremely rare if the ethnic groups have not been engaged in rebellion activities in the past. For all three types o f military repressions, 97-100% of the total instances o f military repressions by the state authorities were against the ethnic groups that had been already involved in rebel activities in the past (see Table 7.5). Surely the data is limited to the most recent period from 1996 to 2000. Recall however that 161

(almost 50%) out o f 338 ethnic groups that are included in the MAR dataset remained peaceful throughout the period covered by the dataset. If the state repressions simply occur at equal chance for all ethnic groups, nearly half repressions would have been directed against the ethnic group that has not been involved in the rebellion activities. Given this, the extremely rare occurrence of military repressions against an ethnic group that has not been involved in rebellion is quite remarkable.

Table 7.5 Previous Involvement in the Rebel Activities by the Targeted Ethnic Group, 1996-2000 No involvement The etlinic group Involved in the in the rebellion started the rebellion rebellion in the Total in the past in tins year past 1 5 0 197 202 Rep 20 2 0 0 18 18 O 3 0 0 Total 5(2 3%) 0 (0°o) 217 (97 7%) 1 0 0 14 14 Rep 21 2 0 0 12 12 3 0 0 16 16 Total 0(0° o) 0(0° o) 42 (100V) 42 1 0 0 1 1 Rep 22 2 1 0 24 25 3 0 0 21 21 Total 1 (2 IV ) 0 (0°o) 46 (97.9V) 47

202 It is understandable that military repressions tend to follow the occurrence of lower-intensity rebellion. As far as an ethnic group remains peaceful, state authorities probably do not find a reason to use military force against the ethnic group, because it would be less costly for the state authorities to maintain the status quo without military operations than to use military force against the ethnic group. The story is different once some members within the ethnic group have taken up arms against the state. There are many reasons why the state authorities would prefer to use force to repress it even when rebellion is not so intense. The state authorities might be afraid of the “demonstration effect” of the rebellion to other members of the same ethnic group (or members of other ethnic groups, if the state is multi-ethnic): the state authorities may try to establish a reputation by using force in the earlier phase of rebellion in order to deter potential rebels. The onset of rebellion, in addition, might foster ethnic antagonism and mistrust in the country and lead to the increased support within the dominant ethnic group for tough measures against the rebels. In this case, the state authorities may be encouraged to take tough measures to gain more support from the dominant ethnic group. Chapter 9 will show, for example, that

Milosevic decided to take tough measures against the KLA rebels in order to re-establish the legitimacy of his regime, and that some ethnic Macedonians were radicalised after the onset of the NLA rebellion and violent demonstrations were held against President Trajkovski who took a conciliatory stance towards the rebels.

Conclusion

This chapter has examined the dynamic relationship between ethnic groups and state authorities. The first part of the chapter presented theoretical

203 arguments about the dynamic relationship between the lower-intensity rebellion, state repressions and large-scale rebellion. It presented the main hypothesis that the occurrence of lower-intensity rebellion incites military repressions by the state authorities, and in turn, these repressions lead to the escalation of rebellion. The second part of this chapter then conducted a series of large-N analyses, using the MAR dataset, to test these hypotheses. The empirical results support the main hypothesis of this chapter, while there is much weaker evidence that structural conditions affect the level of military repressions or the escalation of rebellion after the initial onset.

In order to examine the dynamics in real-life situations, the next chapter will conduct a comparative case-study of Serbia and Macedonia regarding the occurrence of lower-intensity rebellion, state repressions and the escalation of rebellion.

204 Chapter 8 Repression or Conciliation? Comparative Case Study of Serbia and Macedonia

Introduction

This chapter conducts a comparative case-study of two ex-Yugoslav cases, Kosovo (Serbia) and Macedonia. As discussed in Chapter 2, these two cases are different in terms of the course of conflict after the onset of initial rebellion: while both countries experienced an onset of a low-intensity rebellion, Serbia experienced an escalation of rebellion whereas Macedonia did not. This chapter attempts to analyse and explain this difference between

Serbia and Macedonia.

The previous chapter argued that the state repression is the key intervening factor which explains the (non-)occurrence of the escalation of rebellion. In line with the argument and findings in the previous chapter, this chapter will show that the nature of state reactions was indeed a key factor that made a difference between Serbia and Macedonia. This chapter firstly examines the state reactions in Serbia and their consequences. It then examines the reactions by the state authorities and the course of conflict in Macedonia.

This chapter concludes with a brief recapitulation of the key findings.

1. Case of Serbia: Military Repression and the Escalation

(1) State Reactions to the Initial Rebellion

As discussed in Chapter 6, decisions to take up arms made by the KLA leaders were affected by a number of factors, such as a series of discriminatory

205 policies, growing disillusionment of the non-violent strategy of the LDK, and the availability of a large number of weapons. These factors, however, are not sufficient to explain the escalation of rebellion in Kosovo. According to Judah

(2000a: 129), despite the availability of arms and the general frustration of the population, “KLA activists were still finding it hard to win widespread acceptance for their idea of a general uprising... Jashar Salihu found that few people were prepared to take the guns he was offering them from Albania.”

Even when the core members of the KLA decided to take up arms and started the low-intensity rebellion activities, therefore, the escalation was not an automatic consequence of the onset of rebellion.

What, then, explains the escalation of rebellion in Kosovo? Indeed, the reactions of state authorities played a critical role here. In face of sporadic attacks by the KLA, the Serbian authorities started police operations against them. The first-phase operations took place in early 1998, and these operations were conducted mainly by the MUP forces for two reasons. Firstly, since the security threat posed by the KLA was essentially internal, and there was no foreign invasion which would lead to the legal justification of mobilisation of the VJ forces, it was easier and more natural to mobilise the MUP forces.

Secondly, the MUP forces including Special Police Units and Specialist

Anti-terrorist Units were fully subordinated and loyal to Milosevic, and thus

Milosevid trusted them more than the VJ forces (Gow 2003: 79-89). There were some indications of preparations for the operations. For example, in

December 1997, the Serbian Supreme Defence Council decided “to set an absolute priority on Kosovo” and to strengthen the three army corpses in

PriStina, Leskovac, and NiS.1 In January 1998, increased activities of Serbian

1 NaSaBorba, 1998/01/13, cited in Troebst 1998: 13.

206 security forces and army were noted in Kosovo.2 According to Gow, the MUP forces were deployed in full-combat mode in Kosovo by October 1997, with

MUP and VJ units in reserve deployed in the field around the border of Kosovo inside Serbia proper.3 Despite these indications, however, it seems that decisions to go to strong-hand solutions against the KLA were made secretly.

According to Lukic and Simic (2001: 12), for example, “when the clan leader

Adem Jashari was liquidated in March 1998, the news came as a thunderbolt to many people.”

The police operations started in the Drenica region, where the KLA was reported to have a strong presence. According to Troebst (1998: 14), heavily armed police units were mobilised in the operations, equipped with 20 helicopter gunships and 30 armoured personnel carriers. Between 28 February and 7 March 1998, Serbian police forces attacked villages of Likoshan, Qirez and Prekaz i Poshtem in the Drenica region. One of the main targets was the home compound of Adem Jashari, a local strongman who allegedly killed

Serbian policemen several years ago and was sentenced to 20 years' imprisonment for “terrorism” (AI 1998: 78; Judah 2001: 22-23). In Prekaz, the

Serbian police killed almost all members of the Jashari clan, a total of 59 people.4 While Albanians seem to have offered some armed resistance against the police, many of the killed were civilians: according to Amnesty

International (1998: 18), “it was clear that many of the victims - who included at least 12 women and 11 children - had no involvement in the attacks on

2 Nedeljna NaSa Borba, 1998/01/31-02/01, cited in Troebst 1998: 13. 3 Gow 2003: 201. Gow argues that the decision must have been taken at the political level in February 1997 to have reached this level of readiness by October. See Gow 2003: 201-202. 4 For the list of all people killed in Prekaz during the Serbian police operations, see Tahiri 2006: 86-88. This event was called “Drenica Massacre” by Albanians, and the site of the “massacre,” with ruined buildings of the former Jashari compound, turned into a “memorial complex.” For pictures and descriptions of the complex, see Thafi 2004.

207 police.” In Likoshan and Qirez, the Serbian forces killed 26 Albanians, using machine guns, rocket-propelled grenades, helicopters and armoured vehicles in the operation. Similar acts followed thereafter: for example, in May 1998, at least eight men, two of them over 60 years of age, were extra-judicially executed by the Serbian police in the village of Lybeniq, and eight men from

Paklek near Gllogoc were “disappeared” after being detained by police (AI 1998).

The events in Lybeniq show how the Serbian authorities responded to the guerrilla activities in an excessive manner.5 On 25 May 1998, a civilian car was shot at close to the village of Lybeniq, and three men in the car were hit, including a police officer and an off-duty reserve police officer. Terrified that their village would be attacked in retaliation, some of the ethnic Albanian men from Lybeniq reportedly tried to pass a message to the police that there were no arms or armed men in the village. There was, however, no chance for any message of this kind to be sent, and a large police force arrived at Lybeniq soon, firing on the Albanian part of the village with artillery and other weapons. By the time the police arrived, most villagers had fled to the nearby woods to become internally displaced persons, and the police killed some people who were running. This shows that the Serbian authorities responded to actions apparently taken by the rebels in an excessive and indiscriminate manner. By the end of May, 300 people were estimated to have been killed since the start of police operations in February (IICK 2000: 72).

(2) Effect of Repression: Expansion of the KLA

The effect of these police operations was, however, was not to pacify but to “electrify” Kosovo (Judah 2001: 23): since the deaths of Jasharis, the

5 The description of the events in this paragraph is taken from AI 1998: 20.

208 KLA began to expand, village militias began to form and clan elders decreed that now was the time to fight the Serbs. According to Shukri Buja, who became a KLA commander in Nerodime Operational Zone in June 1998, about

30 people of the KLA including him returned from Switzerland to Kosovo via

Albania with light weaponry after the events in Prekaz and started recruiting people for the KLA and transporting arms into Kosovo.6 Baton Haxhiu, who was then a journalist of Koha Ditore newspaper in Kosovo, reported that he didn't see more than 100 KLA soldiers in 1997, but “following the Prekaz and

Likoshan massacres, the revolt spread out all over Kosova and everyone thought it could find solutions by taking to the mountains and they rallied around this armed group which was called KLA.”7 According to Hockenos (2003: 247), these “massacre” events became “the rallying cry that would radicalize the Albanian population in Kosovo and the diaspora.” Surroi told that the events in Prekaz were “fundamental moment” and the war in Kosovo started “symbolically” on 5 March 1998, while the KLA emerged publicly earlier.8 In June 1998, even a Serbian police official admitted that the police operation had increased the support for the KLA in the villages away from the main road.9

As a result, the size of the KLA skyrocketed after the events in Prekaz.

Pleurat Sejdiu recalls, for example, that the events in Prekaz “led to a big influx of volunteers” and “it was unstoppable” (Judah 2000a: 141). Shukri Buja told that the KLA “didn't have any problem recruiting soldiers, because there was a great flood of young people to become members of the KLA.”10 It seems that the KLA themselves were surprised by the course of the events. While

6 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 6302 (2002/06/05) 7 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 5428 (2002/05/23). 8 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 3492 (2002/04/19). 9 Observer, 1998/06/14. 10 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 6427-8 (2002/06/06).

209 “everyone was shocked by what was happening, no one more so than KLA men themselves” (Judah 2000a: 140). According to Sejdiu, the plan of the KLA was to start a war in 1999 but they were forced to speed things up due to the large influx of volunteers (Judah 2000a: 141). Kelmendi told as follows: “as of

March 1998, we wanted only partisan-type guerrilla attacks and still did not want to make open rebellion, because Serbia was too powerful. After Prekaz, however, general uprising started against our will and we had to coordinate.”11

Volunteers came also from abroad. For example, in April 1999, a political representative of the KLA in Tirana told: “Until now, the number of people coming from the West, mostly from Germany and Switzerland, has reached eight thousand.”12 Serbian scholars cite the information that the KLA had about 1,200 members in May 1998, but it had grown to 25,000 members by

July 1998.13

One should note that the expansion of the KLA took place in a rather chaotic manner. Besides the original KLA which started the rebellion, a new structure under Bukoshi called the “Armed Forces of the Republic of Kosova

(FARK)” appeared on the ground, as well as local armed fighters organised by youths on a village-by-village level, calling themselves “Rugova’s KLA,” who fought till the end of the war under the impression that the KLA was under

Rugova’s control (Kola 2003: 336-337). Many people also point out that the

KLA had a de-centralised and localised structure and the “general headquarter” was not functioning. For example, according to Kelmendi, “the General

Headquarter of the KLA was not functioning until very late... it was during the

Rambouillet conference that the zone commanders sat down and elected

Sylejman Selimi as the KLA general commander, whom Agim Qeku replaced

11 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30. 12 AIM Press, 1999/04/20. 13 Mijalkovski & Damjanov 2002: 128.

210 in May 1999 when President Tudjman finally allowed him to leave the post in

Croatia and let him go to Kosovo.”14 Ramush Haradinaj, who was a zone commander of Dukagjini area during the conflict, also said that “there was no general commander” until the selection of Sylejman Selimi.15 Shemsi Syla, who was then a zone deputy commander of the Karadak area, explained as follows: “General Headquarters of the KLA did not give orders... they just gave

‘advices’ and worked on ‘coordination’ between zone commanders because we were waging a guerrilla war and decisions had to be made by zone commanders according to the specifics of Operative Zones.”16

In any case, the quick expansion of the KLA naturally led to the escalation of the intensity of armed conflict in Kosovo. For example, General

Delid testified that there was a sharp increase in the number of attacks by the

KLA since March 199817: he stated that in January and February 1998 the situation in Kosovo remained as usual with only a few attacks, but then “from the month of March onwards, there was a particularly large number of attacks against civilians, a large number of attacks against members of the MUP,” and as a result, there were only about 70 attacks in 1997 whereas in 1998 there were 1470.18 In this way, the KLA quickly expanded. According to the British

14 Interview with Ibrahim Kelmendi, Tetovo, 2006/05/30. 15 Hamzaj 2000: 128. Haradinaj also explains how he was elected as a zone commander in a meeting of representatives from all headquarters of villages in Gllogjan. See Hamzaj 2000: 71-72. 16 Interview with Shemsi Syla, Prishtina, 2006/05/10. According to him, the KLA had seven “operative zones (OZs)” in Kosovo, each of which had its zone commander. These were as follows: OZ Drenica (zone commander: Sami Lushtaku), OZ Pashtrik (Musa Jashari, Ekrem Rexha and Tahir Sinani), OZ Dukagjini (Ramush Haradinaj), OZ Shala (Rahman Rama), OZ Llap (Rustem Mustafa), OZ Nerodime (Shukri Buja), and OZ Karadak (Ahmet Isufi). See also RS 2003, map 1, for the territorial division of the operative zones of the KLA. 1 He did so, of course, without admitting that the sharp increase in the number of KLA attacks was a result of the excessive use of force by the Serbian security forces. Indeed, he denied any excessive use of force by the Serbian authorities. See ICTY transcript of case IT-02-54 (Slobodan Milosevic): 9327-9329 (2005/06/22). 18 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 41242-41243 (2005/06/21).

211 colonel John Crosland, who was a defence attache to Belgrade from 1996 to

1999, the KLA had taken control of three major routes in Kosovo by the summer 1998.19

(3) Second-Phase Operations: Serbian Summer Offensive

Some Serbian political leaders were concerned with the escalation of the conflict and inappropriate conduct by the police forces. According to Lukic and Simic (2001: 12-15), a meeting was held in June 1998 at the Presidential

Palace of Serbia, where political, military and security elite of the country gathered. When the Serbian Minister of Internal Affairs Vlajko Stojiljikovic presented his idea that “all Albanians have to be defeated,” Federal Vice-Prime Minister Lilic heckled him and said: “because of what some elements of your police have done in Kosovo, one day we will be ashamed of being Serb. Our children will be ashamed as well, and maybe even children of our children”

(Luki6 & Simid 2001: 14). However, Stojiljikovic enjoyed support from Mira Markovic (Milosevic’s wife) and Lilic’s objection could not change the repressive course taken by the Serbian government. On 17 June 1998, based on

Milosevic’s proposal, the Serbian Supreme Defence Council decided that “if the terrorist activity of the Albanian separatist movements escalates, the VJ will intervene adequately” (Lukic & Simid 2001: 15-16).

In July 1998, the Serbian authorities started the second-phase operations against the KLA. This time, the VJ forces were mobilised and participated in the operations with the MUP forces. On 23 July 1998, a commandant of the Pristina Corps of the VJ Nebojsa Pavkovic issued a document to the commandants of the 3rd army, emphasising the necessity “to wipe out terrorism in Kosovo” and telling that “regardless of the activity of

19 ICTY transcript of case IT-02-54 (Slobodan Milo§evic): 7926 (2002/07/10).

212 international factors, the VJ has a task and obligation to protect integrity of the country which is now most directly threatened.” According to colonel

Crosland, the participation of the VJ was justified by the expansion of the border area for which the VJ forces were responsible: under the constitution of

Yugoslavia, the sole task of the VJ was to defend the borders of Yugoslavia and the VJ was allowed to maintain control over the borders to 500 metres, but that was then further enlarged to two kilometres and eventually out to 20 kilometres.21 Artillery of various calibres, battle tanks and armoured personnel carriers of the VJ were mobilised, which provided the heavy indirect and direct fire support to the MUP forces.22 Tens of thousands of Albanians fled to hide in the hills and the woods, the houses were looted and burned, and the crops and cattle were destroyed. By the beginning of August 1998, reports estimated that between 200,000 and 300,000 Albanians had been displaced from their homes as a result of military operations (13CK 2000: 74).

Some of the top cadre in the security apparatus were against these operations. For example, the Chief of the General Staff of the VJ Momcilo

PeriSic wrote a letter to Milosevic in July 1998, in which he expressed his concern about the “negative facts” such as the “constant tendency to use the VJ outside the system’s institutions” and “the attempt by unauthorised persons to command VJ units.” 23 He publicly expressed his dissatisfaction over

Milosevic’s policy towards Kosovo in October 1998, telling the Serbian newspaper Blic that Yugoslavia was in the impossible position of being at war

20 The entire text of the document is presented in Mijalkovski & Damjanov 2002: 469-471. 21 ICTY transcript of case IT-02-54 (Slobodan MiloSevic): 7926,7931 (2002/07/10). 22 ICTY transcript of case IT-02-54 (Slobodan MiloSevi6 ): 7927 (2002/07/10). 23 A photocopy of the letter is presented in Lukic & Simic 2001: 160-164. The content of the letter has been presented by the Prosecutor in the trial of Milosevic at the Hague tribunal for several times. See, for example, ICTY transcript of case IT-02-54 (Slobodan Milosevic): 4936-4942. (2002/05/15).

213 against the world and no country can afford to be in that position (Gow 2003:

73). Perisic’s view was also supported by the head of the Serbian Security

Service Jovica StaniSic, who had been regarded by many people as the second or third most important decision maker in the MiloSevic regime by the mid-1990s, and Milorad Vucelic, a politician who was associated with “the pragmatic national wing” in the Milosevic’s party, the Socialist Party of Serbia

(Socijalisticka partija Srbije, SPS hereafter), but both StaniSi6 and VuSelid were removed in October 1998 (Cohen 2002: 300-303). General PeriSic was also sacked in November 1998 and replaced by General Dragoljub Ojdanic, who was a supporter of the leftist party, Yugoslav Left (Jugoslovenska levica,

JUL hereafter) led by Mira Markovic.24 Due to Milosevic’s determination on repressive measures against the rebels and his firm control over the personnel, the opposition to repressive measures within the regime could not moderate the regime’s reactions against the rebellion. The Serbian military campaign of the summer of 1998 was “in many ways a success” (IICK 2000: 75). The KLA had been effectively uprooted: the

KLA “simply melted into the woods,” realising that it could not take on the far more heavily armed Serbs (Judah 2001: 23). But as more and more people fled home to become displaced persons, which was extensively covered by the international media, international pressures on Milosevic to halt military operations mounted.25 On 23 September 1998 the Security Council adopted

Resolution 1199 calling for a cease-fire, the withdrawal of security forces, and cooperation with the international monitoring efforts, and NATO approved an

24 Gow 2003: 72-75. Peri§ic claimed that he was dismissed “in an inadequate and illegal way.” See Cohen 2002: 303-304. 55 Hockenos argues that it was a result of the KLA’s strategy. He argues that “part of the KLA strategy entailed provoking the Serbs, getting them to lash back with predictable ferocity and thus forcing a Western military response” because “the more cruel the repression, the more vivid the message that Albanians could not live under Serb domination.” See Hockenos 2003: 250.

214 “activation warning” for air campaigns against Yugoslavia.26 At the end of

September 1998, attempting to avoid being attacked by NATO, Serbian Prime

Minister Mirko Maijanovic in a special session of the parliament said that

“peace reigns in Kosovo... As of today all anti-terrorist activities have ended.

They will be renewed only if any new bandit and terrorist activity reappears,” even though some operations against rebels were still being conducted.27 In any case, these claims did not impress the Western countries and the US special envoy Richard Holbrooke was dispatched to Belgrade to negotiate with

MiloSevic. Threatened with the prospect of NATO air strikes, Milosevic made an agreement with Holbrooke on 13 October 1998. By the end of October, some 4,000 Serbian special police forces had been withdrawn and the OSCE

Kosovo Verification Mission (KVM), a team of 2000 observers who would monitor enforcement of the agreement, was deployed.

(4) From the Holbrooke Agreement to Rambouillet This agreement did not lead to peace in Kosovo, since the KLA took advantage of the withdrawal of Serbian forces and renewed its military actions.

Haradinaj admitted that the Serbian offensive in the summer had inflicted heavy losses on the KLA and called the Holbrooke agreement “life saving for the KLA” (Hamzaj 2000: 115). Agim Qeku, who became the KLA Chief of

Staff during the 1999 war, later said that the ceasefire was very useful for them

(Sell 2002: 293). During the period with a reduced level of fighting, KLA members returned to burnt-out villages easily since Serbs did not have enough men to hold on to territories they had seized from the KLA, and the KLA had

26 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 6966-7 (2002/06/13). 27 Guardian, 1998/09/28;Financial Times, 1998/09/29. This claim was also repeated by Zivadin Jovanovic, Yugoslav foreign minister, speaking at the UN in New York. See, Financial Times, 1998/09/30.

215 “the time to train seriously and to consolidate a rather chaotic command structure” (Judah 2001: 24). While the UN, NATO and OSCE were alarmed by the KLA's actions and called upon the KLA to cease provocative actions, the

KLA continued its military activities. A series of events in December marked serious deterioration. Facing the KLA military activities, the Serbian army again moved into Kosovo in large numbers, with tanks and other heavy military equipments, establishing permanent positions in various areas in

Kosovo.

The failure of the KVM to prevent an escalation of conflict left an important lesson: in order to control the situation and to prevent escalation, mediators must talk to both sides. General Naumann, the chairman of NATO Military Committee from 1996 to 1999, admitted that many of the incidents were triggered by the KLA while the KVM was operating in Kosovo.

According to him, NATO was not allowed to talk to any KLA man or representative, because NATO initially called them “terrorists” and the iron rule was that NATO will never negotiate with terrorists. He said, however, that

“with hindsight, it was perhaps a mistake,” because “if you are in such a conflict, you should presumably talk to both sides and try to influence them to the best - in the best possible way” and they “could have reduced difficulties in this area” if they talked to and negotiated with the KLA.28

The final key event took place on 15 January 1999, when Serbian forces assaulted Racak village and executed 45 ethnic Albanians. The

OSCE-KVM investigated the site on the next day and concluded that they found “evidence of arbitrary detentions, extra-judicial killings and mutilation of unarmed civilians” (IICK 2000: 81), while the Serbian authorities denied that any civilians had been killed. Both the Security Council and the OSCE

28 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 6995-7 (2002/06/13).

216 strongly condemned the “massacre” at Racak as well as the subsequent Serbian refusal to permit the ICTY access to Kosovo, calling for the immediate cessation of hostilities and dialogue. Following up on these calls for dialogue, the Contact Group composed of six countries, namely the US, Russia, Britain,

France, Germany and Italy, organised peace negotiations to be held in

Rambouillet, France. When the peace talks failed due to the refusal of the

Yugoslav delegation to sign, NATO started air strikes on 24 March, which lasted for 78 days. After the NATO bombardment started, Serbian forces started even harsher attacks on the ethnic Albanians. During the NATO air campaign, it is estimated that almost 90% of the total Albanian population were displaced from their homes and the death toll is estimated to be around 10,000, with the vast majority of victims being Albanians killed by Serbian forces

(IICK 2000: 90-91). When Milosevic finally accepted the G8 principles,

Serbian forces withdrew from Kosovo and the UN administration in Kosovo was established, which effectively ended insurrections by the KLA.

2. Case of Macedonia: Peace Process and Prevention of Escalation

Despite the onset of low-intensity rebellion, Macedonia did not see a serious escalation of conflict, at least when one considers the number of casualties: while around 10,000 people are estimated to have been killed in the case of Kosovo, it is estimated that between 150 and 250 people were killed in

Macedonia in the six months of fighting between the Macedonian security forces and ethnic Albanian guerrillas (Phillips 2004: 161). Therefore, it is clear that the armed conflict in Macedonia was far short of the level of severity of the Kosovo conflict in terms of casualties. As discussed in Chapter 4, if one applies a common operationalisation of civil war which takes 1,000 deaths as

217 the threshold of the “onset” of civil war, the Macedonian conflict would clearly not be classified as a “civil war.” This section will show that one important factor that explains the prevention of escalation was the nature of reactions by the state authorities: contrary to the case of Serbia, the state authorities in

Macedonia did not resort to military repression and launched a political process to peacefully solve the conflict. The key figure in the state authorities in this regard was the then President of Macedonia, Boris Trajkovski.

(1) Trajkovski’s Political Process for the Solution of Conflict

The first effort by Trajkovski in launching a political process for the peaceful solution of the conflict was to set up a national unity government in which all parties, including the Albanian ones, would participate to foster a political dialogue. On 2 April 2001, Trajkovski convened the first meeting of representatives of all political parties to address inter-ethnic issues (Daskalovski 2005: 91). At the fifth round of the all-party talks, the four major parties agreed to form a “government of national unity” (Daskalovski 2005:

92). It was eventually formed in May 2001.29

This coalition, however, went into deep disarray when two Albanian parties and the NLA signed “Prizren Declaration” at the end of May, which stated that Albanian leaders, mindful of an historic juncture in Macedonia, agreed to act in the national interest towards a common goal (Phillips 2004:

118; Ordanoski 2004; Rusi 2004: 8). While it was a part of the peace plan conceived by Robert Frowick, the former American diplomat and then special representative of OSCE in Skopje, it was opposed even by key representatives of the international community, such as Mark Dickinson, then British

Ambassador to Macedonia, who at the time was also representing the EU High

29 Guardian, 2001/05/09; Guardian, 2001/05/14. See also Balalovska et al 2002: 28-30.

218 Representative for Common Foreign and Security Policies, Javier Solana (Rusi

2004: 9). Both Macedonian politicians and the public viewed it as “treason” of the Albanian party leaders against Macedonia (Ordanoski 2004: 21). For example, the daily newspaper Nova Makedonija wrote that “Xhaferi and Imeni signed a document betraying Macedonia.” (Balalovska et al 2002: 31). Both

Georgievski, the then Prime Minister of Macedonia, and Trajkovski strongly opposed such a deal. The dominant feeling in the government was that the

VMRO-DPMNE had been betrayed by its coalition partner the DPA, and

Georgievski stated that “the agreement represents a declaration of war by the

Albanians against the Macedonian nation” (Rusi 2004: 9). Trajkovski also stated that “these meetings are unacceptable” (Rusi 2004: 9) and met Xhaferi and Imeri to demand that they denounce the deal, while they refused to do so, saying it offered the only chance for peace.30 At this point, even Trajkovski ruled out negotiation with the rebels.

Trajkovski continued his efforts to peacefully solve the conflict, however. On 8 June, Trajkovski presented Parliament with a “crisis resolution plan” based on three foundations, namely “the intensifying of political dialogue, the establishment of a special militaiy-police force, and amnesty for people that were forced or enticed to take up weapons.”31 This plan was adopted by the government on 12 June (Daskalovski 2005: 94). Phillips (2004: 121) argues that this action was important because the government had previously been unwilling to consider taking any concrete steps to induce guerrilla defections.

When some members of the Macedonian Academy of Sciences and Arts proposed a partition of Macedonia and an exchange of territories and population between Macedonia and Albania as the only solution, which

30 Guardian, 2001/05/29. 31 Trajkovski 2005: 53-54. On this peace plan, see also Daskalovski 2005: 94-95; Balalovska et al 2002: 34-35.

219 Georgievski tacitly supported, Trajkovski's office responded promptly by publishing a statement that ‘"the only way to resolve these problems is through dialogue with the legitimately elected representatives of the Albanians and their parties and by developing life in common” (Phillips 2004: 124). After the

Macedonian forces failed to retake Aracinovo in June, thousands of enraged nationalists chanting anti-Albanian slogans demanded the resignation of

Trajkovski, stormed the presidential palace, fired shots in the air and attacked foreigners (Phillips 2004: 128-129). While some even feared that Trajkovski might resign, he insisted that a political solution to meet the demands of

Albanians was the only viable option. His determination to solve the conflict peacefully was of critical importance for the continuation of the peace process in Macedonia.

(2) Negotiations at Ohrid While the political dialogue between Macedonian and Albanian parties led by Trajkovski progressed in an inconsistent manner and faced a deadlock, it was announced that the political dialogue was back on track at the end of July.

The venue for dialogue was moved out of Skopje. At first, Trajkovski announced that the next round of discussions would take place in Tetovo, the second largest city in Macedonia with a large Albanian population, “because this town needs peace and we should show that it is a safe town and a symbol of coexistence between different ethnic groups” (Popetrevski & LatifI 2004:

31). This did not materialise due to security concerns, however, and the town of

Ohrid was chosen. The Ohrid negotiations began on 28 July.

The Ohrid agreement was a result of difficult negotiations and compromises from both sides. The issues that caused a deadlock were solved

220 one by one.32 Firstly they solved the language issue. Under the terms of agreement, the Macedonian language is the official language in Macedonia, used for international relations. Any other language spoken by at least 20 percent of the population is also an official language, and can be used for personal documents, civil and criminal proceedings, and so on. Albanian was thus granted this recognition, but the recognition came as a function of demographics. Secondly, a compromise was reached on police reform. It was agreed that the chief of the local police department would be chosen by the municipal parliament from a shortlist provided by the minister of interior and that the ethnic composition of the police force in each municipality should mirror that of the overall population in Macedonia, while Albanians initially demanded that it should mirror the ethnic composition of the municipality. The third issue was the timing of disarmament. The VMRO-DPMNE demanded that rebels disarm before the parliamentary procedure to adopt the constitutional changes. However, it was finally agreed that the disarmament and parliamentary debate should proceed at the same time.

The final obstacle was the continuing violence on the ground. The special anti-terrorist unit of the Ministry of Interior conducted an operation and killed five armed Albanians on 7 August. While all parties accepted the peace agreement on 8 August, an ambush of the NLA killed 10 Macedonian soldiers on the same day and 8 more soldiers were killed when their army vehicle hit an anti-tank landmine. Macedonian forces then started a major offensive, fuelling a cycle of revenge killings, which led to one of the most serious civilian casualties of the Albanian population. The regular and reservist police forces of the Ministry of Interior shelled the village of Ljuboten on 10-11 August and

For the details of the negotiation process at Ohrid, see Popetrevski & Latifi 2004. The descriptions of the negotiation in this paragraph are based on their work.

221 •3*5 conducted house-to-house attacks in the village on 12 August. According to the Human Rights Watch (2001), the Macedonian police forces indiscriminately shelled the village, causing the deaths of civilians including a six-year-old boy and a sixty-six-year-old man. This shadowed the prospects for the peaceful solution of the conflict. For example, one participant in the peace negotiations said that “the atmosphere was tense and difficult,” and “everybody withdrew to their camps, nobody talked to anybody about anything, let alone thought about signing the agreement” (Popetrevski & Latifi 2004: 35). Even when all parties finally agreed to sign the document on 13 August, the location of the signing ceremony was not disclosed until the very last moment due to security concerns. In the end, the Ohrid Framework Agreement was signed at the presidential residence on 13 August 2001. It clearly paved the way for the prevention of further escalation. As a part of the Ohrid peace agreement, NATO authorised the deployment of their forces to disarm the NLA and started “Operation Essential Harvest,” and most of the NLA rebels handed over weapons to the NATO forces. Even though some rebels refused to disarm and formed the “Albanian National Liberation Army” to continue the armed struggle,34 violence largely ceased after the agreement. Vlado Popovski, the then defence secretary, said in April 2002 that the atmosphere between

Macedonians and Albanians was good and they would not fight against each other again.35 The NLA leader Ali Ahmeti said that “If the Ohrid Agreement had not been signed, a new Bosnia would have unfolded with one or two hundred thousand victims” (Popetrevski & Latifi 2004: 36).

33 For the details of the operations, see HRW 2001. 34 For example, a minor clash occurred between former NLA soldiers and the fighters of Albanian National Liberation Army in March 2002. SeeRFE/RL Newsline, 6-57, Part II, 2002/03/26. 35 RFE/RL Newsline, 6-72,2002/04/17.

222 (3) Importance of Trajkovski’s Political Process

How significant was the role of Trajkovski in the prevention of conflict escalation in Macedonia? Some local observers suggest, at least implicitly, that the results would have been the same even if there was no

Trajkovski who led and promoted political dialogue between the Macedonian and Albanian parties. Professor Biljana Vankovska, for example, argues that

“Macedonians are not militant...we have no militant heroic history. People were not enthusiastic for being mobilised to the army and avoided call-up for mobilisation. Otherwise, it would have been another Bosnia.”36 If the presence or lack of militant history fully explains the presence or absence of escalation into the large-scale conflict, it would mean that results would have been the same even if there was no Trajkovski who took an initiative to start the peace process. This thesis argues, however, that the conflict indeed could have been escalated into a large-scale one if there had not been a peace process led by

Trajkovski, because other leading politicians in the government such as Prime

Minister Georgievski and Minister of Interior BoSkoski showed more radical attitudes towards the Albanian rebels. For example, when Trajkovski was making his efforts to form a “government of national unity,” Georgievski pressed for a declaration of war, arguing that it would lead to a more efficient fight against guerrillas (Balalovska et al 2002: 29). When Trajkovski announced a peace plan at the beginning of June, the VMRO-DPMNE criticised it, saying “we call on some officials to stop misinforming the

Macedonian public about the difficult security situation by their allegedly peaceful initiatives and demands for amnesty of terrorists, whose victims are

36 Interview with Biljana Vankovska, Skopje, 2006/05/29.

223 members of the Macedonian security forces” (Balalovska et al 2002: 35). The

Prime Minister’s spokesman told that the Pardew-Leotard proposal to make

Albanian the second official language was a “shameless” document which would “seal the doom of the Republic of Macedonia” (Balalovska et al 2002: 37).

BoSkoski was even more radical. For BoSkoski, the elimination of the

“terrorists” was a “holy task” (Balalovska et al 2002: 29), and he was reported to be personally present in Ljuboten during the entire operation on August 12 when the Macedonian police forces conducted a house-to-house attack.37 He created two paramilitary groups, “Tigers” who were aggressive police unit technically under the control of the Macedonian police commander, and

“Lions” who were an unauthorized body of former police and military reservists until the autumn 2001, when they became an official police unit.38 In

January 2002, the independent magazine Forum reported that the Lions included members with “renowned criminal history” and that it was created not to defeat the Albanian extremists but to help in achieving other political goals, namely escalation of military clashes and the elimination of those who oppose the radical military solution proposed by BoSkoski (Phillips 2004: 158). The case of Ljuboten shows that these radical attitudes were not just a political gesture to win popularity among ethnic Macedonians but with real intentions.

If there had been no leading politician to start and foster the peace process, therefore, the cycle of violence between the state authorities and the rebels might have escalated into a large-scale conflict.

37 HRW 2001. Boskoski was prosecuted by the Hague Tribunal for “violations of the laws or customs of war” in Ljuboten. See the indictment against Ljube Boskoski and Johan Tarculovski, Case No. IT-04-82, available at http://www.un.org/ictv/indictment/english/ bos-ii050309e.htm 38 Phillips 2004: 157. For details of the Tigers and Lions, see also Ordanoski 2004.

224 Conclusion

This chapter examined the responses of the state authorities against initial rebellions and their effects on the course of conflict. It firstly analysed the case of Kosovo, where the Serbian authorities decided to take harsh repressive measures against the KLA. These measures mobilised heavily armed police and military forces and led to indiscriminate killings of civilians. These repressive measures led to a sharp increase of the members of the KLA, and thus led to an escalation of rebellion. This chapter then analysed the case of

Macedonia, where Trajkovski started a peace process at a relatively early phase of the conflict. Despite the difficulties and deadlocks, this peace process successfully found a political solution to the demands of the Albanian rebels and led to the cessation of armed conflict before the rebellion escalated into a large-scale conflict. The different responses by the state authorities explain the occurrence or prevention (lack) of escalation of rebellion in Serbia and

Macedonia. These findings from the comparative case-study in this chapter, therefore, confirm the theoretical argument and the empirical findings in the previous chapter.

The conclusion in this chapter raises one fundamental question: why, then, did the state authorities respond so differently to the onset of initial rebellion? What were the factors that affected the choice and decisions made by the state authorities? These questions will be addressed in the following chapter.

225 Chapter 9 Explaining State Reactions: Path Dependency and External Pressures in Serbia and Macedonia

Introduction

The previous chapter has shown that the state authorities in Serbia and

Macedonia reacted to the onset of rebellion in a different manner, which determined the occurrence or non-occurrence of escalation. This chapter conducts an additional comparative case-study on the determinants of the state reactions in the two countries.

This chapter is structured as follows. The first section will argue that the factors considered in Chapter 7 are not sufficient to explain the difference between Serbia and Macedonia regarding the reactions to the initial rebellion.

Other factors emphasised in the existing literature also fail to explain it. This section will thus introduce two explanatory factors, namely (1) path dependency of the policy position taken by key politicians towards ethnic minorities and (2) external pressures. The second section will analyse the

Serbian case and show that the path-dependency of policy position taken by

Milo§evic towards Albanians, reinforced by the domestic political context, and the ambiguous signals from the international community led to a decision by the Serbian government to take repressive measures against the KLA. The third section then analyses the Macedonian case and shows that Trajkovski was fairly moderate since his inauguration and that external actors put pressure more effectively on the Macedonian government to restrain its reactions to the rebels.

226 1. Explanatory Factors: Path Dependency and External Pressures

This chapter uses two explanatory factors that were not used in

Chapter 7. They are necessary because the factors examined in Chapter 7 are not sufficient to explain the difference between Serbia and Macedonia. The key

explanatory variable that explains state repression in Chapter 7 was the

occurrence of low-intensity rebellion. However, this factor cannot explain the

difference since both countries experienced the occurrence of a low-intensity rebellion. As for the economic development, the level of economic development was 50% higher in Serbia than Macedonia as of 1988 (PleStina

1992: 180-181). While the lack of data in Yugoslavia hinders the precise comparison in the 1990s, the difference between Serbia and Macedonia in terms of economic development would not be so large that it could explain the different state reactions in Serbia and Macedonia.1 The nature of the political regime is one factor that may explain the difference between Serbia and Macedonia: while Serbia had an authoritarian regime (polity2 score is -6 for

Serbia in 1998), the political regime in Macedonia was relatively democratic

(polity2 score is 6 for Macedonia from 1993 to 2000). This factor, however, cannot explain why Trajkovski took a moderate stance towards rebels while other politicians, such as Georgievski and Boskoski, took a more radical position, since all of them were democratically elected in Macedonia. The nature of the political regime, therefore, cannot fully explain the moderate position taken by Trajkovski.

Other factors that are used in the existing literature to explain different levels of state violence also fail to explain the difference between Serbia and

1 In the Penn World Tables 6 .1 used in the present research, the data of GDP per capita in Yugoslavia is missing for the entire period from 1990 to 2000. See Penn World Tables, httn://dc2 . chass.utoronto. ca/nwt/alphacountries .html

227 Macedonia. For example, Ron (2003) argues that “institutional settings”

(whether the area is “frontier” or “ghetto” for the state authorities) affect the level of violence employed by the state authorities. According to him, Serbia resorted to ethnic cleansing (i.e. extremely high level of state violence) because

Bosnia was “frontier” and external to Serbia’s core, and thus the state authorities did not feel a bureaucratic, moral and political sense of responsibility for its fate, whereas Israeli state violence in Palestine was more restrained because it was “ghetto” and inside the Israeli core, and thus the state authorities felt more responsibility (Ron 2003: 8-9, 13-24). Ron argues that the restraint shown by the Serbian authorities in the Sandzak area, which is inside the Serbian core, also supports his argument. As for Kosovo, he argues that the Serbian authorities employed the tactics of ethnic-policing until 1999, and they started ethnic cleansing operations because Kosovo’s institutional setting had been dramatically transformed: due to the NATO intervention, Kosovo’s ghetto status evaporated and became “externalized,” and this led to the ethnic cleansing operations by the Serbian authorities (Ron 2003: 87-111).

His argument, however, cannot explain the difference between Serbia and Macedonia. According to his own criteria, both Kosovo and north-western

Macedonia were inside the “core” of the state when the state authorities made their reactions to the initial rebellion (note that the Serbian police operations started in early 1998, when Kosovo remained inside the “core” of the Serbian state according to Ron). Despite the similarity between two cases in terms of the “institutional settings” discussed by Ron, the state authorities in the two countries showed a remarkable difference in terms of their reactions to the initial rebellion.

In order to explain the different policy choices made by Milosevic and

Trajkovski, this chapter examines two additional factors. The first is the

228 path-dependency of policy choices. Widely used in political science as well as in economics and other fields, the concept of “path dependency” means that the current outcomes depend on the path of previous outcomes, rather than simply on current conditions, and thus “history matters.”2 This concept is used to explain the policy choices as well. For example, Schmidt (2002) argues that one factor that affects public policy besides institutions is the path dependency of public policy, i.e. the feedback coming from policy choices and outcomes in the past. The present chapter argues that the logic of path dependency explains policy choices made by Milosevic and Trajkovski. The second factor is the external pressure, which often limits policy choices made by the governments and politicians. The present chapter will argue that the Western countries, especially the U.S., sent a wrong signal to MiloSevid, which was interpreted as a “green-light” for repression, while they put more coherent pressure on the

Macedonian government not to resort to repressive measures.

2. Case of Serbia

(1) Path-dependency and the Context of Domestic Politics

Why did Milosevic decide to take repressive measures against the

KLA rather than a more conciliatory stance? In order to explain the policy choice made by Milosevic, one needs to go back to the origin of his ascent to power in the late 1980s and see what his agenda was. Indeed, it was precisely the issue of Kosovo that Milosevic used to foster nationalism as his power-base.

For example, in April 1987 in Kosovo Polje, Milosevic told the audience: “the process of emigration of Serbs and Montenegrins under the economic, political

For the usage of the concept of path dependency in political science, see, e.g., Pierson 2000; Collier & Collier 1991.

229 or constant physical pressures constitutes probably the last tragic exodus of

European population... You must stay here. This is your land” (Milosevic 1989:

144). In conclusion, he said: “entire Yugoslavia is with you... Yugoslavia does not exist without Kosovo! Yugoslavia disintegrates without Kosovo!

Yugoslavia and Serbia would not give Kosovo away!” (Milosevic 1989: 146).

On 28 June 1989, at the Gazimestan shrine which memorialises the 1389 Battle of Kosovo, Milosevic made a famous speech to commemorate 600th anniversary of the Battle. When Milosevic walked to the platform, he was greeted by shouts “Slobo! Slobo!” and “Kosovo is Serbia!” (Cohen 2002: 144).

In his speech, he criticised the “concessions many Serbian leaders made at the expense of their people” as historically and ethically unacceptable, and declared that “Serbia is united today and equal to other republics,”3 apparently praising the amendments of the Serbian constitution which effectively annulled the autonomous status of Kosovo.

Kosovo thus had a critical political importance for Milosevic, which made it difficult, or even impossible, for him to make concessions on the issue of Kosovo. A number of Western diplomats recalled that Milosevi6 became difficult whenever the issue of Kosovo was raised. For example, former US ambassador to Yugoslavia Warren Zimmermann (1996: 57) wrote: “Kosovo was Milosevic's hottest button. He was unyielding, emotional, pugnacious, and full of invective for its Albanian inhabitants.” Lord Owen (1995: 137) recalled that “[o]ver Kosovo the polite mask sometimes broke and we would be into an ugly confrontation. It was as if he [Milosevic] knew this was the area of his most indefensible behaviour on which he was personally vulnerable, and he would sometimes turn snarling on me or anyone who raised it.” General Clark

(2001: 65) wrote as follows: “During the shuttle discussions and at Dayton,

3 Borba, 1989/06/29.

230 Dick Holbrooke and Chris Hill tried several times to raise the issue of Kosovo with Milosevic. No deal. ‘This is internal matter for Serb people and

Albanians,’ Milosevic said. He stood his ground.” Louis Sell (2002: 281-282), a Foreign Service officer of the US Department of State, observed that the

Serbian control over Kosovo was “virtually all that he had left to show for a decade of disastrous rule” by the end of the 1990s and “Milosevic could not afford to give up Kosovo voluntarily, even if he had wanted to.” According to him, MiloSevi6 told a US diplomat in early 1998 as follows: “Kosovo is not

Bosnia. Kosovo is my head” (Sell 2002: 281).

Here, one can see the logic of path dependency of policy position. It was MiloSevid who raised the issue of Kosovo and called for “unity” as the solution of the problem, and the realisation of such “unity” (meaning the abolition of autonomy in Kosovo) was loudly publicised as a major achievement of his government. Precisely due to this position taken in the past, giving up the “unity” of Serbia by granting Kosovo even autonomy, let alone full independence, was clearly political suicide for Milosevic. According to

Sell (2002: 281), this political importance of Kosovo for him was overlooked by the US diplomats who attempted to achieve peace through negotiations with

Milosevic: Washington's response to Kosovo, he observed, was complicated by a “lingering nostalgia that the conflict in Kosovo, like Bosnia, could be resolved by dealing with Milosevic.” He argued that “Kosovo’s significance to his own survival meant that Milosevic was prepared to go to war with the most powerful military alliance on the planet, something that few seem to have expected” (Sell 2002: 283).

This logic of path dependency was reinforced by the context of domestic politics, namely the declining legitimacy of the regime. Cohen (2002:

267-271) points out that Milosevic and the SPS were condemned by many

231 Serbs for having “betrayed” Serbian national interests, and that Milosevic could no longer portray himself as a heroic patriot in the second half of the

1990s. The decreasing support for Milosevic and declining legitimacy of his regime in 1996 and 1997 has been well documented. According to Sekelj

(2000: 72), for example, public opinion polls showed that Milosevic’s popularity had dropped from active support of more than half the population in

1992 to only 12% before the 1997 elections. According to a survey more than half of the citizens expressed no-confidence in the political institutions (such as

President, Government and Parliament) as early as 1996,4 and confidence in political institutions dropped further in 1997.5 For example, only 29% expressed confidence in President of Serbia (i.e. Milosevic) and 61% expressed no-confidence in 1997 (Slavujevid 1999: 33-37). Slavujevic (1999: 100) concluded that the Serbian political system enjoyed the support of only one-fifth of citizens and was completely de-legitimized in 1997.

As a result, serious challenges were posed to the regime in 1996 and 1997. Firstly, the reformist coalition Zajedno won in major cities including the capital, Belgrade, in the 1996 local elections (Thomas 1999: 285). When the election commission refused to recognise the results, the opposition coalition waged huge demonstrations, which lasted for more than two months (Thomas

1999: 285-318). In February 1997, the regime was forced to admit the victory of the opposition and the Zajedno coalition assumed power. Secondly, in

Montenegro, reformist Milo Djukanovic ran as a presidential candidate against incumbent Momir Bulatovic, Milosevic’s close ally, in the 1997 presidential elections, and the former won (Thomas 1999: 379-386, Goati 2001: 138-148).

Electoral victories of the Zajedno coalition in Serbia and Djukanovic in

4 As for the details of the survey results, see Slavujevic 1997. 5 As for the details of the survey results, see Slavujevic 1999.

232 Montenegro posed serious challenges to the MiloSevic regime.

Furthermore, even nationalists in Serbia, who had been core supporters of the Milosevic regime, became unsatisfied with him and started shifting their support to the Serbian Radical Party {Srpska radikalna stranka ,

SRS hereafter), led by ultra-nationalist Vojislav Seselj. In the 1997 elections,

SeSelj was able to “tap the deep sense of national humiliation” (Cohen 2002:

269) by emphasising his commitment to Serbia's ethno-national aspirations, which many Serbs felt Milosevic had betrayed. Helped also by the boycott of some opposition parties that constituted the Zajedno coalition, the SRS won

29.3% of the vote and 82 out of total 250 seats in the Serbian Parliament, and

SeSelj even defeated the SPS candidate for the Serbian President winning 49.1% of the votes, even though the overall turnout did not reach 50% and thus the election was declared invalid.6 Some local observers argue that the growing support for Seselj was not a real threat for Milosevic, because “they were the same”7 and “SeSelj was merely a junior partner of Milosevic.”8 The growing support for Seselj and the SRS, however, was clearly a signal to

MiloSevic that even nationalists were not satisfied with his regime any more.9

In this context of declining legitimacy, it was even more dangerous for

MiloSevic to take a conciliatory stance on Kosovo, which would further discredit him in the eyes of nationalists but would not bring him any additional support from reformists. Therefore, MiloSevic chose to re-establish his legitimacy by taking a firm stance toward Kosovo. For example, on 10

December 1997, the Yugoslav delegations walked out of the Bonn conference

6 SRS officials protested that the authorities had manipulated the turn-out figures in order to deprive SeSelj of his victory. See, Thomas 1999: 352. Eventually another candidate from the SPS, Milutinovic, defeated SeSelj in the presidential election in December 1997 and the SPS managed to remain in power. 7 Interview with Sonja Biserko, Belgrade, 2006/04/06. 8 Interview with Dejan Anastasijevic, Belgrade, 2006/04/26. 9 Interview with Tank Begic, Prishtina, 2006/05/05.

233 on the implementation of the Dayton agreement when an attempt was made to place Kosovo on the agenda: in the run-up to the 21 December presidential elections, the SPS had to appear to be taking a firm line on Kosovo (Simic

2000: 179; Thomas 1999: 403). Furthermore, Milosevic decided to co-opt

Seselj into the ruling coalition of the Serbian government exactly when he started an offensive against Albanian “terrorists”: on 24 March 1998, a new government was formed by a so-called “patriotic coalition” of the SPS, the JUL and the SRS (Simic 2000: 103); the SRS received 16 of 36 portfolios, with

SeSelj taking the post of deputy prime minister. As Thomas (1999:418) argued,

“the entry of the Radicals into positions of power meant that there would be no moderation in the government’s stance on the issue of Kosovo.” Some analysts even speculated that Milosevic's decision to launch an offensive in Kosovo may have been influenced by his desire to create a psychology of acute threat: while any rapprochement between the ruling parties and the SRS had appeared impossible after the December 1997 elections, a new atmosphere of crisis over

Kosovo brought the two camps closer and thus weakened criticism against the regime (Thomas 1999: 416). Mihailovic (1999: 145) also argues that the

Serbian elite invented a legitimacy formula by activating a crisis in Kosovo.

In April 1998, MiloSevic organised a referendum in Serbia, in which the voters were asked whether they “accept the participation of foreign representatives in the resolution of problem in Kosovo and Metohija,”10 and overwhelming majority of voters voted for “no” (Cohen 2002: 286). When he announced his proposal of the referendum, MiloSevic said: “we have rejected to accept the participation of foreign representatives in the resolution of internal question of our country, and especially in the resolution of problem in

10 Politika, 1998/04/03.

234 Kosovo and Metohija.”11 This referendum was used to bolster the popular support for his regime as well as to “democratically” legitimise his tough position against Albanians. His decision to take a firm stance towards Albanian

“terrorists” and to use force against them was thus clearly a vital component of his strategy to re-establish the legitimacy of his regime.

(2) Ambiguous Signals from External Actors In addition to the path dependency and political context, ambiguous signals sent from the external actors also encouraged the Serbian authorities to take repressive measures. The behaviour of Robert Gelbard, then US special envoy for the Balkans, is often mentioned in this regard. When he made his first visit to Kosovo in January 1998, Gelbard was quoted in Serbian newspapers as saying that Washington was considering putting the KLA on its list of terrorist organisations barred entry to the US.12 Importantly, already in

January 1998, the US was well aware of the preparations for police operations in Kosovo. Madeleine Albright (2004: 380), the then US Secretary of State, recalls that “[f]rom sources in the region, we received word in January 1998 that MiloSevic was preparing to respond with a military crackdown.” This awareness notwithstanding, Gelbard apparently did not publicly express his opposition to repressive operations. In February 1998, immediately before the

Serbian police operations in Kosovo, Gelbard condemned attacks carried out by the KLA, which he described as a “terrorist organisation.”13 In addition, the

US even made some concessions toward Yugoslavia at the end of February

1998: Gelbard told MiloSevic that Washington was prepared to let Yugoslavia join the South-east European Co-operative Initiative, a US-backed body

11 Politika, 1998/04/03. 12 Financial limes, 1998/01/09. 13 Financial Times, 1998/02/24; The Times 1998/03/02; Sunday Times 1998/03/01.

235 promoting economic development in the region.14 Yugoslavia would also be allowed to increase its diplomatic representation to its UN observer mission, to open a consulate in the US and to apply for landing rights in the US for charter flights by its national airline.15

These actions sent a signal to the Serbian authorities that the repressive measures against the KLA were justifiable. For example, General

Deli6 and Milosevic quoted Gelbard’s remarks to support their claim that the

KLA was a “terrorist” organisation.16 Cohen (2002: 283) argued that

“inadvertently perhaps, Gelbard had provided Milosevic with a green light to launch his March 1998 offensive in Kosovo. Indeed, Serbian action in Kosovo escalated only four days after Gelbard’s remarks.” Simic (2000: 194) observed that Gelbard’s condemnation of the KLA as a terrorist organisation was interpreted as “signal for action” by the Serbian government circles. Petritsch and Pichler also pointed out that “Gelbard indirectly supported Milosevic's firm course... It was not by chance that the attack followed immediately after the visit of Gelbard...For MiloSevic, Gelbard's signal meant a green light for military solution” (PetriC & Pihler 2002: 82).

This does not mean, however, that the US did not attempt at all to send a message that it was against the use of force in Kosovo. The US had been saying that there must be significant progress on Kosovo before it would lift

“outer wall” sanctions on Yugoslavia.17 At the end of February, Gelbard urged

MiloSevic not to launch a military crackdown in Kosovo and to open talks with

14 Financial Times, 1998/02/24. 15 Financial Times, 1998/02/24. Note however that these concessions were made primarily in return for the Belgrade’s support for a new pro-Western Bosnian Serb government and were not officially linked to any actions related to the situation in Kosovo. 16 ICTY transcript of case IT-02-54 (Slobodan Milosevic): 4266 (2002/05/03), 41248 (2005/06/21) 7 FinancialTimes, 1998/01/09.

236 Rugova.18 After the reports of massacre at the beginning of March 1998,

Gelbard said that the US will not tolerate violence and that violence will lead to the toughest consequences imaginable.19 Gelbard met MiloSevic after the

Prekaz massacre and told him that “you have done more than anyone to increase membership in the KLA. You are acting as if you are their secret membership chairman,” which infuriated MiloSevic (Sell 2002: 281). Gelbard flew to Kosovo on 10 March and condemned the Serbian violence, which helped to restore his standing in Kosovo.20 The US officials thus certainly attempted to send a signal to MiloSevic to discourage the use of force against

Albanians in Kosovo. However, their attempts were blurred by their ambivalent attitudes towards the KLA: even after the Prekaz massacre, US officials still continued calling the KLA “terrorists.” In May 1998, for example,

US envoys Holbrooke and Gelbard again condemned the KLA as terrorist and attempted to persuade Rugova to do so, even though he refused.21

The attitude of the US clearly changed in June in favour of the KLA, apparently due to the continuing atrocities in Kosovo. At the beginning of June, it was reported that US officials were against British proposals to seal the mountainous border between Kosovo and Albania, because the US was concerned that arms may be denied to the KLA.22 In mid-June, NATO started an air exercise code-named “Determined Falcon” in Macedonia and Albania to

“demonstrate NATO’s capability to project power rapidly into the region.”23

The decisive event occurred at the end of June: Holbrooke talked to the KLA members for the first time in the border town of Junik.24 While Holbrooke

18 Financial Times, 1998/03/02. 19 Independent, 1998/03/04. 20 Times, 1998/03/11. 21 Financial Times, 1998/05/11; 1998/05/12; 1998/05/13. 22 Observer, 1998/06/07. 23 Observer, 1998/06/14. 24 Times, 1998/06/25.

237 claimed that the meeting was “unplanned” and “unofficial,” this meeting was

reported as a clear signal to Milosevic that they cannot be dismissed as

“terrorists” any more.25 Only a few days later, Holbrooke revealed that

Gelbard had made official contact with the KLA representatives in Geneva,26 and the US even suggested that the KLA should participate in the peace talks with Belgrade.27 Albright (2004: 286) recalls as follows: “By mid June, it had become obvious that no political settlement would be possible without the rebels, so our diplomats began meeting with the KLA representatives. This infuriated MiloSevic and nettled the Europeans, but it was the only way to make progress.”

Even after the Western officials stopped condemning the KLA as “terrorists,” MiloSevic was not deterred from using force. One may argue that the diplomatic pressure from outside generally had a limited effect on the

Serbian regime, since MiloSevic had “rarely been bothered about international criticism before.”28 In addition, however, the division in the major powers also made the external pressure ineffective. The most important in this regard was the Russian factor : while Russia participated in the Contact Group and attempted to broker a peace agreement, it consistently opposed military intervention against Serbia. For example, when Western officials started discussing military intervention as a “last resort,” Russia emphasised that it would strongly oppose any move to invade Yugoslav territory. When NATO started the air exercise “Determined Falcon,” Milosevic visited Moscow and

25 Guardian, 1998/06/26. 26 Financial Times, 1998/06/30. 27 Guardian, 1998/06/30. 28 Independent, 1998/03/25. 29 As for the good review of “Russian Factor” in the Balkans since the end of Cold War, see Simic 2000: 137-161. For the relationship between NATO and Russia during the Kosovo crisis, see also Norris 2005. 30 Sunday Times, 1998/06/07.

238 held a meeting with President Boris Yeltsin, where MiloSevic received an assurance that Russia would veto NATO’s military intervention in the UN

Security Council.31 The Western countries were also not always monolithic: for example, when the US suggested that the KLA should be included in the peace talks, EU foreign ministers were divided over such a suggestion. When the US started taking tougher stance against MiloSevid, it was supported by the

UK and Germany but less so by France and Italy.33 As Wolff (2003: 87) pointed out, a “major problem that inhibited the international community’s ability to devise and implement effective conflict prevention, management and resolution policies resulted from the fact that there was no unified approach to the Kosovo crisis.”

3. Case of Macedonia

Contrary to MiloSevic in Serbia, Trajkovski took a conciliatory stance towards the rebels and played an important role in achieving a peace agreement in August 2001. This section will analyse why he could, and did, play such a positive role in Macedonia.

(1) Path Dependency: Election with the Support of Albanians

A crucial difference between Milosevic and Trajkovski was the origin of their ascent to power and the paths chosen by the two Presidents at the time of their inauguration. MiloSevic gained power in Serbia precisely by waging

31 PetriC & Pihler 2002: 95-96. This Milosevic-Yeltsin agreement led to the establishment of KDOM (Kosovo Diplomatic Observer Mission), which turned into the KVM (Kosovo Verification Mission) after the Milosevic-Holbrooke agreement in October 1998. See PetriC & Pihler 2002: 96-97. 32 Guardian, 1998/06/30. 33 Financial Times, 1998/04/29. See also Albright 2004: 381-382 for the attitude o f France, Italy and Russia.

239 nationalistic campaigns on Kosovo. On the contrary, Trajkovski was elected because he was more moderate towards Albanians, not because he was radical and nationalistic.

Table 9.1 Results of the Presidential Elections in Macedonia, 1999 Fust Roimd Second Round Registered voters 1.613.284 Votes cast 1.050.499 Votes Cast 1.120,087 j Voter Turnout 65.11% Voter Turnout 69 56% Invalid Votes 11.013 Invalid Votes: 17.271 Candidates Votes % Candidates Votes 0 o Tito Petkovski 343,606 32.70% Boris Trajkovski 591,972 52 80% Boris Trajkovski 219,098 20 90% Tito Petkovski 514,599 45 90% Vasil Tupurkovski 163.206 15.50% Muarem Nexipi 155,978 14 80% Stojan Andov 111,983 10 70% Muliamed Halili 45,731 4 40% Source: Website of the Election Commission of Macedonia http \nnv izbori.gov. ink'

The electoral system played a crucial role for the election of

Trajkovski. In Macedonia, the two-ballot majority system is used for presidential elections, in which the second round of voting is fought only by the two leading candidates in the first round vote (assuming that no candidate achieved a majority on the first ballot). In the 1999 presidential elections,

Petkovski was leading in the first round vote while Trajkovski ranked second, but Trajkovski won in the second round (Table 9.1). Albanian voters had a decisive role in determining the winner: they voted en masse for Trajkovski in the second round because his position was more moderate towards Albanians than Petkovski. According to Xhaferi, he suggested to Albanian voters that they should vote for Trajkovski in the second round “because he was new, atypical

240 (Protestant rather than Orthodox), young and moderate,”34 and they followed

his suggestion in the second round.

One can clearly see the massive shift of votes from Albanian

candidates to Trajkovski by comparing the number of votes for two Albanian

candidates in the first round and the increase in the number of votes for

Trajkovski in the second round (see Appendix 6). The number of votes

received by the two Albanian candidates is positively correlated with the

increase of votes received by Trajkovski: Pearson’s correlation coefficient

between these two figures is extremely high (0.976) and statistically significant

at the 0.01 level (two-tailed test).35 It is thus clear that votes received by two

Albanian candidates shifted massively to Trajkovski in the second round. The number of votes received by two Albanian candidates exceeded 200,000 in the

first round, which is more than half of the additional votes received by

Trajkovski in the second round. Clearly, Trajkovski could not have won the

elections if there were no support from Albanians in the second round.

It was thus only natural that Trajkovski delivered a moderate message

regarding inter-ethnic relations in Macedonia. In his inauguration speech, for

example, he said: “throughout the past decade, the Republic of Macedonia has

often been taken as an example for interethnic tolerance. I will not allow ethnic

hatred, jingoism and intolerance to threaten the stability of the country. The

ethnic groups represent a patrimony and the Republic of Macedonia must have a strategic and permanent orientation to preserve, nourish and develop them.”

34 Interview with Arben Xhaferi, Tetovo, 2006/05/30. Xhaferi was then the president o f the DPA, the largest Albanian party which was in the ruling coalition with die VMRO- DPMNE since 1998. Therefore, his recommendation was highly influential with Albanian voters. The multivariate regression analysis confirms that this association remains valid even when other factors (change of the total votes cast between two rounds, number o f votes for Andov and Tupurkovski in the first round vote) are controlled for. Results of the multivariate regression analysis are not shown but available from the author upon request. 36 Inauguration speech o f President Trajkovski, December 15 1999. For the English

241 Did the fact that Albanians voted for him in the presidential elections

affect the decisions and actions made by Trajkovski in reacting to the

occurrence of rebellion? At least some people seem to have hoped so in the

early phase of rebellion. For example, former President of Albania Berisha called for moderate reactions of the Macedonian authorities toward Albanians,

emphasising that he would like “to remind Mr. Trajkovski that the Albanians have voted for him.”37 However, those who were close to Trajkovski argue that his religious belief was the primary factor: Robert Milcev, a brother-in-law of Trajkovski, and Jason Miko, who worked closely with Trajkovski for the government’s lobbying activities in the US, both said that it was the “Christian perspective that it is better to talk than to fight” that affected his decisions and actions during the armed conflict.38

In any case, Trajkovski was moderate since his inauguration, and it was natural for him to take an initiative to foster political dialogue between

Macedonian and Albanian parties, which led to the Ohrid peace agreement and de-escalation of the conflict. While some local scholars doubt that Trajkovski played an active role in the peace process,39 many people assert that he played an active role in making a peace plan. For example, Mark Laity, a NATO press officer who was in the cabinet of Trajkovski during the period of conflict, said that “the president has created a big plan to end the conflict, a kind of road map that combined his instincts and ideas and International Community ideas”

translation of the original text, see http://www.boristraikovski.org/ 37 AIM Press, 2001/03/28. Interview with Robert Milcev and Jason Miko, Skopje, 2006/06/01. Trajkovski was a devout Methodist and religiously very active since his youth. For example, in 1978, when he was 22 years old, he was elected as president of the Evangelical Methodist Youth of Yugoslavia. See Miko 2004. •5Q For example, Professor Biljana Vankovska said that “President Trajkovski certainly was a nice, decent man, man of peace and very responsive to the international community, but he was not of a high intellectual capability... peace plan was made by Americans, because it could not go without them.” Interview with Biljana Vankovska, Skopje, 2006/05/29.

242 (Phillips 2004: 121). Jason Miko reported that “the idea of the peace plan to

bring all political parties to talk and negotiate was primarily made by

Trajkovski and his cabinet, especially his national security advisor, Nikola

Dimitrov.”40 Muharem Nexhipi, an ethnic Albanian and the then deputy health

minister, said that he believed that Trajkovski was the only politician who

could organise the constitutional change long desired by Albanians (Phillips

2004: 122). The former Foreign Minister Ljubomir Fr5koski, who was an

advisor of Trajkovski during the conflict and an expert participant at the Ohrid negotiations, pointed out that “Trajkovski played a crucial role for the peaceful

solution of conflict, because he was the only politician who could take the responsibility to take care of negotiations. No other Macedonian politicians wanted to lead the peace negotiations with Albanian parties, because they thought they would lose the elections if they did so.”41

(2) External Pressures from the International Community The original policy position of Trajkovski is not the sole factor to explain the reactions of the Macedonian authorities. International factors also played a critically important role. In the case of Macedonia, signals sent from the external actors were more coherent. Until the Ohrid agreement, Western officials kept putting strong pressure on the Macedonian authorities to restrain the use of force against the rebels and to find a political solution to the conflict.

When the NLA rebellion started, external actors such as the EU, the

US and NATO continuously put pressure on the Macedonian authorities so that their reactions would remain proportional to the level of threat and would not become excessive. For example, when the Macedonian forces launched an

40 Interview with Jason Miko, Skopje, 2006/06/01. 41 Interview with Ljubomir FrCkoski, Skopje, 2006/06/07.

243 offensive against the NLA around Tetovo at the end of March 2001, NATO

Secretaiy-General George Robertson and the EU High Representative Javier

Solana arrived in Skopje to put pressure on the Macedonian authorities to halt the offensive (Phillips 2004: 97). After the March offensive, international monitors complained to the government about the arrest and beating of

Albanian civilians and the destruction of houses by security forces.42 At the end of June, when the Macedonian forces started an offensive against the village of Aracinovo, which constituted the strongest action against the NLA and mobilised helicopters, tanks and heavy artillery, Robertson called this action “sheer madness,” demanded its suspension and visited Skopje to persuade the Macedonian state leaders to agree to a ceasefire.43 Eventually,

NATO and OSCE intervened to implement the pull-out of the NLA rebels, together with their arms, and to transfer them to an undisclosed location.44

It was argued above that labelling the KLA as “terrorists” had sent a wrong signal to the Serbian authorities. In this regard, Western officials were much more careful in the case of Macedonia, probably because they learned precisely from their earlier experiences in the Kosovo conflict. When the NLA emerged, Solana condemned the violence and urged Albanian leaders to isolate

“extremists,”45 and NATO also called them an “extremist Albanian group,” 46 but both avoided the word “terrorist.” When Trajkovski visited the US on 1

May 2001, expecting President Bush to openly label the NLA as “terrorists,” the US Report on Terrorism in the World published by the State Department on the same day did not label them as terrorists.47 After the intervention of NATO

42 Guardian, 2001/04/10. 43 AIM Press, 2001/07/01. 44 AIM Press, 2001/07/01. 45 Solana’s statement on the violent incidents in the border region of the former Yugoslav Republic of Macedonia, 2001/03/05, available at http://ue.eu.int/ 46 Daily Telegraph, 2001/03/09. 47 AIM Press, 2001/07/23.

244 and OSCE at Aracinovo, which led to strong anti-West demonstrations by

Macedonians, the US prohibited some people connected to the NLA from entering the US and banned all financial transactions for the NLA from the US, but it was done “just to restore the disturbed balance.”48 Solana used the term

“terrorist” once to condemn the ambush by the NLA which killed eight

Macedonian commandos, but he did not forget to emphasise the importance of a political solution of the conflict, saying that “by remaining steadfastly on the course of dialogue and reform, the responsible political leaders can effectively counter the threat of extremist nationalism and help their country advance on the road to full participation in a prosperous and stable Europe.”49

The Western countries also attempted to put pressure on the

Macedonian government to start a political process to make a deal with Albanian parties, if not with the rebels directly. At the end of March, for example, Solana pressed the Macedonian authorities to move to a political solution of the problem, saying that “[grievances cannot be allowed to fester.

My message to all Macedonians today is that it is in their interests to address urgently the root causes of legitimate grievance, through political and democratic mechanisms.”50 The European countries also exerted pressure on

Prime Minister Georgievski to abandon the idea of declaring a state of war and to form a government of broad or grand coalition.51 On 18 May 2001, Solana appointed Mark Dickinson, Ambassador of the UK in Skopje, as his personal representative in Macedonia, aiming at “maintaining a daily channel of communication with key players” and assisting “in pushing forward the

48 AIM Press, 2001/07/23. 49 Solana’s statement after the killing of 8 soldiers near Tetovo, 2001/04/29, http://ue.eu. int/ Solana’s statement following his visit to Tetovo, 2001/03/27, http://ue.eu.int/ 51 AIM Press, 2001/05/21.

245 political dialogue.”52

The Western officials intensified pressure for a political solution after

June 2001. At the beginning of June, Solana again visited Macedonia twice, which led to the offer of amnesty by Trajkovski.53 When the peace plan created by Trajkovski was made public at the beginning of June, the ministers of EU countries approved this plan and Solana visited Skopje again to express in person support for the plan.54 When negotiations between the leaders of the four major parties started on 14 June based on his peace plan, ambassadors from the US and the UK were “closely monitoring” the talks at Trajkovski's office.55 At the end of June, when European foreign ministers met in

Luxemburg, they expressed their dissatisfaction with progress achieved so far and made the granting of aid to Macedonia conditional upon progress in political dialogue.56 They assessed that the EU mediation in this dialogue was necessary and appointed the former French Defence Minister, Francois Lyotard, as a full-time special envoy in Macedonia.57 Leotard and James Pardew, US President Bush's special envoy, held intensive individual and joint discussions with local political leaders to reach a political agreement. When negotiations faced a deadlock at the end of July, Solana and Robertson flew to Macedonia to provide additional pressure again, saying “95 percent of the agreement has been already agreed and only 5 percent remains to be agreed upon, which is not worth waging a war.”58 Given such a strong and continuous involvement of the

EU and the US, one can argue that the Ohrid agreement would not have been

Solana’s statement on the confirmation of the appointment of Ambassador Dickinson as his personal representative in fYROM, 2001/05/18, http://ue.eu.int/ 53 Guardian, 2001/06/01. 54 AIM Press, 2001/06/14. 55 AIM Press, 2001/06/23. 56 AIM Press, 2001/07/01. 57 AIM Press, 2001/07/01. 58 AIM Press, 2001/08/21.

246 achieved without their efforts of mediation. For example, Western diplomats in

Skopje concurred that Solana’s dedication to the peace process in Macedonia

turned out to be essential (Phillips 2004: 117).

Why was the pressure from the international community more

effective in the case of Macedonia? Firstly, some scholars suggest that it was

effective because it was timely. For example, the former Foreign Minister of

Macedonia Denko Maleski argues that “the West intervened timely to

peacefully solve the conflict, and they did so because of their earlier

experiences in the Balkans.”59 Indeed, the West responded fairly quickly, at a

very early stage of the conflict, and it certainly increased the effectiveness of

external pressure. Secondly, external actors were more unified and coherent

and did not show internal disagreements as they did in the case of Kosovo. The third important factor was the dependency of Macedonia on external support.

The fact that Macedonia was seeking membership of EU and NATO gave them

leverage on the Macedonian leadership. For example, at the beginning of April

2001, Macedonia signed Stabilization and Association Agreement with the EU,

and at the signing ceremony, Prime Minister Georgievski promised to meet the

June deadline for improved relations between Macedonians and Albanians.60

When Roberson attempted to persuade the Macedonian authorities to halt the offensive against Aracinovo, it was rumoured that he achieved it by threatening

sanctions.61 Macedonian politicians were aware that Macedonia needed external assistance and this awareness certainly made them more responsive to the demands of the international community.

59 Interview with Denko Maleski, Skopje, 2006/06/01. 60 Guardian, 2001/04/10. 61 AIM Press,2 m t0 im .

247 Conclusion

This chapter has examined the factors that explain the difference

between Serbia and Macedonia regarding state reactions to the occurrence of

initial rebellions. The first section pointed out that neither the factors

considered in Chapter 7 nor the “institutional settings” emphasised by Ron can

explain it. It then introduced two additional explanatory factors, namely path

dependency of the policy position and external pressures. The second section

showed that the path-dependency of policy position, reinforced by the domestic

political context, and the ambiguous signals from the international community

led to the repressive measures in Serbia. The third section pointed out that Trajkovski was fairly moderate from his inauguration - he was elected as

President precisely because he was perceived as more moderate by Albanians

in Macedonia - and the external actors put timely, coherent and strong

pressures on the Macedonian government to restrain its reactions to the rebellion.

What are the theoretical implications of the analysis in this chapter?

As for the path dependency of policy position, it implies that it would be much

more difficult to take a moderate and conciliatory stance for politicians who had already taken a radical stance on the same issue, even though it does not

mean that politicians would never change their positions on crucial issues. In the case of Serbia, for example, some local politicians seem to have noticed this difficulty. Biljana PlavSic, then President of Republika Srpska, told Solana and General Clark that “the solution to the problem of Kosovo is democracy in

Belgrade” (Clark 2001: 113). Clark (2001: 130) interpreted this remark as follows: “with a democratic government there, peaceful arrangement could be made to address the Albanians’ concerns. With Milosevic [sic] in power, ... a

248 solution would never be possible.”

As for external pressure, this chapter’s analysis gives us some lessons

for the effectiveness of external pressure for the prevention of escalation.

Firstly, the case of Serbia suggests that external actors should be careful when

they condemn the rebels, because a strong condemnation of the rebels may be

interpreted by the state authorities as a “green-light” for repression. External

actors should be especially careful when they use the term “terrorist,” because the usage of this term may be interpreted as a total and unconditional

de-legitimisation of the rebels, which in turn would be interpreted as

unconditional and full legitimisation for repression. Of course, this does not mean that external actors should not condemn rebels at all. In many cases, the rebels should be blamed for their violent actions, and external actors then should condemn rebels to discourage them from resorting to violence. Tapping a right balance is important, since a condemnation of one side may elicit counter-violence from the other side. Secondly, in order to prevent escalation, actions made by external actors must be timely and coherent. External pressures must be timely because stopping the cycle of violence at an early stage is critically important for the prevention of escalation. External pressures must be coherent because internal disagreements among external actors will decrease the effectiveness by undermining the credibility of possible punishment, i.e. by giving an impression that external actors would not be able to take coherent actions to punish the state that takes repressive measures.

This chapter's key theoretical arguments are not tested by large-N analysis, because it is extremely difficult to operationalise and quantify both factors emphasised in this chapter. How to operationalise and measure, for example, the “original policy position” taken by the key politicians in the state authorities? How to operationalise and measure the “timeliness” or

249 “coherence” of the pressures made by external actors? Finding a convincing answer to these methodological questions is not an easy task. This chapter is a part of the efforts for theory-making rather than for theory-testing. The validity of the theoretical arguments of this chapter remains to be checked and tested in future research.

250 Conclusion

The present thesis has completed its theoretical explorations and

empirical analyses of the occurrence of ethnic rebellion. It now concludes with

a summary of key findings and a discussion on the theoretical and

methodological implications for the study of ethnic conflict as well as the

policy implications for conflict prevention.

1. Key Findings of the Thesis

Part II analysed the initial level of rebellion. Chapter 3 presented two

ideal types, rebellion from above and from below, and found that the initial

level of rebellion is more likely to be higher if politicians are involved in the

organisation of rebellion. Chapter 4 conducted a comparative analysis of Serb

and Albanian rebellions in the ex-Yugoslav region and has shown that the involvement of politicians, police and army in the rebellion led to the high

level of the initial intensity of rebellion in Croatia and Bosnia, while Albanian rebellions were “from below,” organised by those who did not have a large amount of resources, and consequently the initial level of rebellion was low in

Kosovo and Macedonia.

Part III examined the onset of initial rebellion. Chapter 5 argued that structural conditions emphasised in the existing literature may not suffice to explain the onset of initial rebellion and argued for the importance of “dynamic grievances.” In other words, when members of the ethnic group keep making peaceful protests for a long time, people may get increasingly frustrated by the fact that they cannot achieve their goals by peaceful means. As a result, some people decide to choose a violent option as an alternative to change the status

251 quo. Empirically, Chapter 5 found that the probability of the occurrence of

rebellion increases as the duration of peaceful protest gets longer. In addition,

some structural conditions associated with the grievances (such as political

discrimination and loss of autonomy) also had a significant effect on the onset

of initial rebellion. Chapter 6 conducted a case study of the four rebellions in

the ex-Yugoslav region and has shown that “dynamic grievances” explain the

timing of the onset of rebellion in Kosovo and Macedonia, even though

structural conditions do explain the onset of rebellion as well, while other

structural conditions, such as the newness of the state, better explain the early

onset of rebellion in Croatia and Bosnia.

Part IV examined an interaction between the ethnic group and the state. Chapter 7 argued that this interaction becomes important when one analyses

the escalation of rebellion after the onset of initial rebellion and found that the

state authorities are more likely to take repressive measures when the

low-intensity rebellion occurs, and the escalation of rebellion is more likely to occur when the state authorities take repressive measures. Chapter 8 compared

the state reactions to the initial rebellion in Serbia and Macedonia and has

shown that the Serbian authorities took repressive measures against the rebels,

which led to the escalation of rebellion, while the Macedonian authorities took

a more conciliatory stance, which led to the de-escalation of rebellion. Chapter

9 has shown that this difference between Serbia and Macedonia is caused by two factors, namely the path dependency of policy position taken by key politicians and the nature of the external pressure. The policy position taken by

Milosevic in the 1980s made it impossible for him to take a conciliatory stance towards Albanian rebels in the 1990s, and the external actors, especially the US, sent a wrong signal to Milosevic by condemning the KLA as “terrorists,” that was interpreted by the Serbian authorities as “green light” for repression. On

252 the other hand, Trajkovski was more moderate towards Albanians since his

inauguration and thus it was natural for him to take a conciliatory stance, and the external actors put a coherent and clear pressure to find a political solution to the conflict.

2. Implications of the Case Selection

The introductory chapter of this thesis discussed the issues related to the case selection and the generalisability of the theoretical arguments derived from the case study. This section elaborates on the implications of the case selection. After the comparative case study and the large-N analysis, can we say that the ex-Yugoslav cases are typical cases of ethnic rebellion? Are they rather anomalies or atypical cases? In what sense are they typical or atypical?

What about the factors excluded from the large-N analysis? This section attempts to address these questions. In what sense are the ex-Yugoslav cases typical or atypical? One can use the results of the large-N analyses to answer this question. If these cases follow the general pattern in the dataset, the ex-Yugoslav cases can be regarded as typical cases, at least regarding the aspect analysed in the thesis, rather than anomalies. The results of the large-N analyses in Chapter 3 suggest that all of the ex-Yugoslav cases are typical in a sense that the involvement of the elite, such as politicians and military officials, led to the high intensity of initial rebellion (Serb rebellions in Croatia and Bosnia) while the decisions from below (made by non-leaders) led to the low intensity of initial rebellion

(Albanian rebellions in Kosovo and Macedonia). Based on the empirical findings in Chapter 5, one can argue that the cases of Albanian rebellions in

Kosovo and Macedonia are typical in a sense that the lack of achievements of

253 the peaceful protest that lasted for a relatively long period led to the decisions

to take up arms by some members of the ethnic group. While the Serb leaders

in Croatia and Bosnia decided to take up arms earlier rather than later, it is not

certain whether they are typical or atypical, because the large-N analyses in

Chapter 5 do not give a definitive answer. As for the findings in Chapter 7, they

suggest that the case of Serbia is typical in that the occurrence of low-intensity

rebellion led to the harsh police and military operations against the rebels, and

these operations led to the escalation of rebellion. One can also argue that the

case of Macedonia is rather atypical because the occurrence of low-intensity

rebellion did not lead to the repressive operations by the state and the

conciliatory stance taken by the state authorities led to the de-escalation of rebellion.

Overall, the large-N analyses tell us more about the typicality of the

ex-Yugoslav cases, but not much about their peculiarities. However, what is

excluded from the large-N analyses does tell us about the peculiarities of the

ex-Yugoslav cases. These peculiarities are not trivial, but are of importance for the further theory-building on ethnic conflict and civil war. This section discusses two such factors, namely (1) availability of arms and (2) role of external actor.

In Part El, Chapter 6 pointed out that the availability of arms played an important role in determining the timing of rebellion in Kosovo and

Macedonia, but this factor was excluded from the large-N analyses in Chapter

5. This factor also played an important role in the case of Serb rebellions in

Croatia and Bosnia in that the availability of a large amount of weapons and other resources led to the decisions by the Serb leaders to take up arms earlier rather than later. This factor is of enormous importance particularly for the conflict prevention: this factor may act as a major limiting factor to prevent a

254 large-scale conflict even when the grievances among the population are high

and all of the “right conditions for insurgency” are met. One should try to

explore possibilities to incorporate this factor into an explanatory framework

and operationalise it adequately to include in the large-N analysis.

This factor also tells us about one of the peculiarities of the

ex-Yugoslav cases. Chapter 4 argued that the local politicians could mobilise a

large number of military resources to organise a rebellion. This, however, is not

always the case across the world. In many countries, the local politicians do not

have any access to the weapons or military resources whatsoever. In the case of

the ex-Yugoslav region, the Territorial Defence system made it possible even

for the local politicians to access the weapons: in order to facilitate the guerrilla

warfare against the foreign invader, a large number of small arms and

ammunition were stored in every municipality in ex-Yugoslavia (see Chapter 4).

This kind of institutional setting is obviously not so common in the world.

While this does not necessarily invalidate the hypotheses drawn from the ex-Yugoslav cases because they are supported by the large-N analyses, this

does suggest that the way in which politicians or elites acquire arms and other

military resources for a rebellion may vary from case to case.

Chapter 9 pointed out that the external actor (international

community) played an important role in determining the state reaction against the rebels, but this factor was not included in the large-N analyses in Chapter 7.

This factor is also connected to the availability of arms discussed above: the

external actors may provide weapons and other resources to the organisers of rebellion. In the ex-Yugoslav cases, the Serbian government and the Federal military authorities (the JNA leaders) provided a considerable amount of political, economic, and military resources for the Serb leaders in Croatia and

Bosnia. The external actors played an important role in the cases outside

255 ex-Yugoslavia as well. During the cold war, many internal conflicts were

fuelled by two superpowers, the US and the Soviet Union, who provided a

considerable amount of financial and military resources for the opposing

parties in the conflict. Laitin (2001a) concluded that the external actor played a

decisive role in the onset of secessionist rebellion in the former Soviet Union.

In the existing literature, the large-N analyses of the onset of rebellion and civil

war have not adequately modelled this factor. As discussed in Chapter 9, it is

extremely difficult to quantitatively measure the effect of the actions made by

the external actors in a methodologically convincing manner. This factor,

however, needs to be included in the explanatory framework in some way or

another.

The ex-Yugoslav cases are also quite unique, and probably atypical, regarding the role of external actors. The ex-Yugoslav conflict attracted

considerable attention of the international media and policy makers, which was certainly disproportionate to the conflicts that occurred in the other places in the 1990s. For example, NATO has never used its military power but against

Serbia to rescue an ethnic minority from the military repression in the name of humanitarian crisis. While the involvement of external actors per se is not necessarily rare in the internal conflict as discussed above, its scale and magnitude in the ex-Yugoslav conflict may be rather exceptional.

Of course, in order to examine to what extent the ex-Yugoslav cases are unique or atypical in a methodologically robust manner, one must conduct a comparative analysis across region and period, comparing them with the cases from other regions and periods. This is beyond the scope of the present thesis and remains a task for future research.

256 3. Theoretical and Methodological Implications

As for the theoretical and methodological implications of the present

thesis, two particularly important issues are discussed here. The first is related

to the unitary actor assumption of the ethnic group. The second is related to the

conceptualisation of the onset of civil war as a one-shot event, or the

operationalisation of the onset of civil war as a binary variable.

(1) Unitary Actor Assumption

The arguments and analyses in Part II have significant implications for

the unitary actor assumption that is often retained in the game-theoretic accounts of ethnic conflict. Ethnic groups do not act as unitary actors in reality,

and there are often disagreements not only on the objectives of ethnic groups but also on the means to achieve such objectives. The present thesis has shown that radicals who advocate a violent means to achieve their goals can emerge both in the powerful ethnic elite and among non-leaders. Whether radicals emerge in the former or the latter will affect the initial intensity of rebellion.

The unitary actor assumption becomes especially problematic in the cases of rebellion caused by decisions from below. If radicals emerge among non-leaders within the ethnic group and thus the initial intensity of rebellion is not so high, the occurrence of large-scale ethnic rebellion cannot be fully explained simply by revealing the factors or dynamics that make the radicals take up arms. If the rebellion starts from a low level, then there must be a process of escalation before it becomes a large-scale rebellion. In this case, ethnic groups do not take up arms en masse , but the number of people who take up arms increases over time. In these circumstances, the unitary actor assumption becomes problematic because any theory based on such an

257 assumption cannot explain why an escalation or de-escalation occurs after the onset of initial rebellion.

The game-theoretic accounts of ethnic rebellion may encounter fewer problems in the cases of rebellion caused by decisions from above. Decisions made by powerful leaders often leads immediately to the occurrence of large-scale rebellion, and thus a primary task for the theories of ethnic conflict is to explain the decisions made by these leaders. Here, the strategic calculations of ethnic leaders may suffice as an explanation of the occurrence of large-scale rebellion. The empirical analysis of Chapter 3 shows, however, that rebellions tend to start from lower levels of intensity than from higher levels: more than 100 out of total 177 initial onsets occur at fairly low intensity, from 1 to 3 on the rebellion index (see Table 3.5). This implies that the usefulness of the game-theoretic accounts based on the unitary actor assumption is limited to a small subset of cases, even if one admits that the unitary actor assumption is reasonable in the cases of rebellion caused by decisions from above.

(2) Conceptualisation and Operationalisation of Civil War

As discussed in Chapter 1, it is a common practice in the existing literature to use a binary coding for the occurrence of civil war. The occurrence of civil war is thus conceptualised as a one-shot event, and the objective conditions observed in the year of the “occurrence” will be associated with the onset of civil war and thus be judged as “causes” of civil war in the statistical analysis.

The findings of the present thesis suggest that this practice can be seriously misleading. In order to illustrate the problem, let us take the example of Sri Lanka. In the dataset of Fearon and Laitin, the “onset” of civil war

258 between the government and the LTTE is coded in 1983. In their explanatory

framework, therefore, the structural conditions observed in Sri Lanka in 1983

are assumed to explain this “onset” of civil war. However, this assumption is

not quite plausible because the LTTE started its low-intensity rebellion in the

early 1970s, the state authorities then took severe repressive measures which

inflicted serious casualties on the Sri Lankan Tamils, including many civilians,

and this state repression radicalised the Tamils and led to the escalation of

rebellion in the early 1980s (e.g. Bose 1994). While both the onset of

low-intensity rebellion and the state repression took place before 1983, these

factors are not considered in the explanatory framework of Fearon and Laitin.

The explanatory framework of the present thesis will treat the “onset” of

large-scale conflict in the 1980s as a joint outcome of the onset of initial

rebellion in the 1970s and the state repression from the late 1970s to the early

1980s. It will suggest, therefore, that factors that led to the initial onset of

rebellion (such as dynamic grievances) and factors that led to the state repression (such as the path dependency of policy position) should be

examined in the analysis of the causes of conflict. As this example illustrates, the findings of this thesis suggest that it is

not enough to examine the structural conditions observed in the year of “onset”

of large-scale conflict. Especially when large-scale conflicts result from the

onset of low-intensity rebellion, the onset of initial rebellion and the escalation

of rebellion should be analysed separately, because these two occur at different times and may have different causes. The three-step approach proposed in the present thesis can offer an alternative, more nuanced explanatory framework for the analysis of the causes of ethnic conflict and civil war.

259 4. Policy Implications

This section briefly examines the policy implications of this thesis.

When the occurrence of ethnic conflict or civil war is conceived as a one-shot

event, the prescribed recipe for conflict prevention will inevitably be based on

all-or-nothing assumptions: if it succeeds, no conflict occurs; if it fails, the

large-scale conflict will occur. The three-step approach of the present thesis

implies, however, that there are at least two steps to prevent the occurrence of

large-scale conflict. The first is to prevent the initial onset of rebellion, to

which most scholars give their attention. The second is to prevent the

escalation of rebellion, even though this step is available only when the initial level of rebellion is low.

The first step is closely related to the controversy in the existing

literature on the role of grievances. As discussed in Chapter 1, there are

basically two camps. One emphasises the importance of opportunities and argues that “grievances do not matter.” This argument implies that the rebels

are rational and opportunistic. The other emphasises the importance of

grievances. This argument implies that the rebels are frustrated. This camp

argues that the occurrence of rebellion cannot be explained without the

grievances that frustrate potential rebels and motivate them to take up arms.

Lickbach (1998) called the former camp “Collective Action (CA) theorists”

and the latter camp “Deprived Actor (DA) theorists”. The difficult question the

analysts often face is succinctly summarised by Lichbach (1998: 31): “are CA theorists too cynical, or DA theorists too naive?”

The answer to this question affects the policy implications for the prevention of ethnic conflict. If rebels are mere opportunists, there is no benefit in listening to the demands of members of the ethnic group, and conflict

260 prevention is possible only by eliminating the structural conditions that favour the rebellion. If rebels are motivated by grievances, on the other hand, it will be necessary to address the grievances of the ethnic group for the prevention of ethnic conflict. Not surprisingly, political actors who want to suppress the rebellion, such as state officials or military officers, tend to make the former claim: rebels are mere opportunists and “terrorists,” whose voice does not deserve any attention. Rebels tend to claim that they are motivated by the grievances and fight for the “just cause.”

The findings of the present thesis reject extreme and simplistic answers from both ends. This thesis has found that both opportunities and grievances do matter for the initial onset of rebellion. In other words, the decision to take up arms is caused neither purely by opportunities nor purely by grievances. This implies that, for the prevention of ethnic conflict, it is necessary to make efforts in two directions simultaneously, one to address the grievances and the other to eliminate opportunities. On the one hand, policies that try to address the grievances of the ethnic group will not be effective if not accompanied by effective and pragmatic policies that deter potential rebels from taking up arms in the future. On the other hand, policies that try to deter potential rebels from taking up arms will not be effective if not accompanied by some measures to address the grievances of the ethnic group.

As for the second step, the present thesis implies that it is important to prevent the state authorities from resorting to excessive repression when the level of rebellion is low, because excessive repression tends to be counterproductive and lead to the escalation of conflict. Of course, the state must take necessary measures against rebels to protect its citizens. However, this cannot be a pretext for conducting indiscriminate repression against the ethnic group. The external actors, including foreign governments and

261 international NGOs, can play an important role in putting pressure upon the state authorities not to resort to excessive repression.

Indeed, this is where the conventional recipe for conflict prevention can go wrong. Some scholars have argued that the large-scale conflict is caused by the weakness of the state. This argument naturally implies that the recipe for conflict prevention is to “strengthen” the state. If this leads to the strengthening of the military and police of the state, however, it may eventually turn out to be counterproductive, because the state authorities may become too confident in their military and police capabilities to suppress rebels and thus resort to repression. Of course, the military and police capabilities of the state should be strong enough to deter potential rebels from taking up arms for mere opportunistic purposes. However, this “strengthening” of the state should not encourage the state authorities to resort to excessive repression.

5. Tasks for Future Research: Additional Implications and Theoretical

Extensions

Finally, this section elaborates on some issues that deserve particular attention for the further theory-building on ethnic conflict and civil war. This thesis questioned the validity of the unitary actor assumption with regard to ethnic groups and explored some consequences of relaxing this assumption.

One possible way to extend this argument further would be to examine the intra-rebel dynamics within the rebel organisations. The rebel organisation may not be monolithic and may well be divided between militants and moderates.

This division and the balance between these two forces may affect, for example, the timing of the de-escalation of rebellion or the end of civil war: the civil war will continue when militants win over moderates, while the civil war will end

262 when moderates win over militant rebels. There have been few efforts to explore these intra-rebel dynamics and its relationship with the evolution of conflict in a systematic manner. Therefore, this topic may deserve particular attention in future research on ethnic conflict and civil war.

The findings of the present thesis also suggest that the nature of state actions against rebels or ethnic groups is of critical importance for the evolution of conflict. The present thesis has shown that repressive measures taken by state authorities are likely to be provoked by the onset of low-intensity rebellion, but are also affected by other factors, such as international pressures and policy positions taken by key politicians in the past.

However, the validity of these arguments should be checked in more systematic analyses of state actions, including large-N analyses. Therefore, further efforts are needed to systematically analyse and explain state actions taken against the rebels. Most scholars who are interested in the occurrence of large-scale conflict have attempted to explain the decisions made by rebels to take up arms, but not so many efforts have been made to explain the decisions made by the state authorities to take repressive measures to crack-down on rebels rather than to take more conciliatory measures. Therefore, it is necessary to “bring the state back in” to the analysis of ethnic conflict and civil war.

Here again, it may be useful to relax the unitary actor assumption.

State authorities are also not always monolithic, and may well be divided between militant hard-liners and conciliatory moderates. For example, Chapter

8 of this thesis shows that there were some conciliatory moderates even in the

Milo§evic regime in Serbia, and there were some militant hard-liners even in

Macedonia. What made a crucial difference between these two countries was that the militants won over the moderates in Serbia while the moderates won over the militants in Macedonia. By theorising these intra-regime dynamics

263 within state authorities, therefore, one may be able to create a useful explanatory framework that explains the state actions against the rebels.

As discussed in the introductory chapter, this thesis only marks a beginning, rather than an end, of the efforts to develop a more dynamic theoretical framework to explain the occurrence of ethnic conflict and civil war.

Further efforts are needed to develop a more comprehensive explanatory framework. This is far beyond the scope of this thesis, and is a task for future research.

264 Reference

Agresti, Alan. 2002. Categorical Data Analysis, 2nd edition, Hoboken: John

Wiley & Sons.

AI (Amnesty International). 1998. Kosovo: the Evidence, London: Amnesty

International.

Albright, Madeleine. 2004. Madam Secretary: A Memoir, London: Pan

Macmillan.

Andjelic, Neven. 2003. Bosnia-Herzegovina: The End o f a Legacy, London:

Frank Cass.

Amautovic, Saud. 1996. IzboriuBosni i Hercegovini'90, Sarajevo: Promocult.

ASRA (Academy of Sciences of the Republic of Albania), ed. 1993. The Truth

on Kosova, Tirana: Encyclopaedia Publishing House.

Balalovska, Kristina, Alessandro Silj & Mario Zucconi. 2002. Crisis in

Macedonia, Rome: Ethnobarometer.

Ballentine, Karen. 2003. “Beyond Greed and Grievance: Reconsidering the

Economic Dynamics of Armed Conflict,” in Ballentine

2003: 259-283.

Ballentine, Karen & Jake Sherman. 2003. The Political Economy o f Armed

Conflict: Beyond Greed and Grievance , Boulder: Lynne Rienner.

Baric, Nikica. 2005. Srpska Pobuna u Hrvatskoj 1990-1995., Zagreb: Golden

Marketing-TehniSka Knjiga.

Barth, Fredrik. ed. 1969. Ethnic Groups and Boundaries: The Social

Organisation o f Cultural Difference, Oslo: Universitetsforlaget.

BDHRL (Bureau of Democracy, Human Rights, and Labor). 2001. Cambodia:

Country Reports on Human Rights Practices, 2000,

http://www. state, gov/g/drl/rls/hrrpt/ 2000/eap/681 .htm

265 Begic, Kasim. 1997. Bosna i Hercegovina od Vanceove misije do Daytonskog

sporazuma, Sarajevo: Bosanska knjiga.

Beissinger, Mark. R. 1998. “Nationalist Violence and the State: Political

Authority and Contentious Repertoires in the Former USSR,”

Comparative Politics, 30-4: 401-422.

Bieber, Florian & Zidas Daskalovski. eds. 2003. Understanding the War in

Kosovo, London: Frank Cass.

Biserko, Sonja. ed. 2004. Milosevic vs Jugoslavia, knjiga I, Beograd:

HelsinSki odbor za ljudska prava u Srbiji.

Bjorgo, Tore & Rob Witte, eds. 1993. Racist Violence in Europe, New York: St.

Martin's. Bojic, Mehmedalija. 2001. Historija Bosne i Bosnjaka (VII-XX vijek),

Sarajevo: TKD SahinpaSic.

Bose, Sumantra. 1994. States, Nations, Sovereignty: Sri Lanka, India and the

Tamil Eelam Movement, Thousand Oaks: Sage Publications. Bose, Sumantra. 2002. Bosnia after Dayton: Nationalist Partition and

International Intervention, London: Hurst.

Bose, Sumantra. 2003. Kashmir: Roots o f Conflict, Paths to Peace, Cambridge:

Harvard University Press.

Bourne, Mike. 2007. Arming Conflict: The Proliferation of Small Arms,

Hampshire: Palgrave Macmillan.

Brass, Paul. 1997. Theft o f an Idol, Princeton: Princeton University Press.

Brown, Michael E. ed. 1993. Ethnic Conflict and International Security,

Princeton: Princeton University Press.

Brown, Michael E. 1997. “The Causes of Internal Conflict: An Overview,” in

Brown et al 1997: 3-25.

Brown, Michael E., Owen R. Cote Jr., Sean M. Lynn-Jones, & Steven E. Miller.

266 eds. 1997. Nationalism and Ethnic Conflict, Cambridge: MIT Press.

Brubaker, Rogers & David D. Laitin. 1998. “Ethnic and Nationalist Violence”

Annual Review of Sociology, 24: 423-452.

Bugajski, Junusz. 1994. Ethnic Politics in Eastern Europe: A Guide to

Nationality Politics, Organisations, and Parties, Armonk: M. E.

Sharpe.

Burg, Steven L. & Paul S. Shoup. 1999. The War in Bosnia-Herzegovina:

Ethnic Conflict and International Intervention, Armonk: M. E.

Sharpe.

Churcher, Bob. 2002. “Kosovo Lindore/Preshevo 1999-2002 & The FYROM

Conflict,” Conflict Studies Research Centre (CSRC) Research Paper, http://www.da.mod.uk/CSRC/documents/balkans/G104

Ciger, Norman. 2001. “Serb War Effort and Termination of the War,” in Magas

&2ani6 2001:200-235.

Clark, Howard. 2000. Civil Resistance in Kosovo, London: Pluto.

Clark, Robert P. 1984. The Basque Insurgents: ETA, 1952-1980, Madison:

University of Wisconsin Press. Clark, Wesley K. 2001. Waging Modern War: Bosnia, Kosovo and the Future

o f Combat, Oxford: Public Affairs. Coakley, John. ed. 2003. The Territorial Management o f Ethnic Conflict,

second and expanded edition, London: Frank Cass.

Cohen, Lenard J. 1993. Broken Bonds: The Disintegration o f Yugoslavia,

Boulder: Westview.

Cohen, Lenard J. 2002. Serpent in the Bosom: The Rise and Fall o f Slobodan

Milosevic, Boulder: Westview.

Collier, Paul. 1999. “Doing Well Out of War,” World Bank Research Paper,

http://econ.worldbank.org/programs/conflict/librarv/

267 Collier, Paul & Anke Hoeffler. 1998. “On Economic Causes of Civil War,”

Oxford Economic Papers, 50-4: 563-573.

Collier, Paul & Anke Hoeffler. 1999. “Justice-Seeking and Loot-Seeking in

Civil War,” World Bank Research Paper, http://econ.worldbank.org/

Collier, Paul & Anke Hoeffler. 2001. “Greed and Grievance in Civil War,”

World Bank Research Paper, http://econ.worldbank.org/

Collier, Ruth Berins & David Collier. 1991. Shaping the Political Arena:

Critical Junctures, the Labor Movement, and Regime Dynamics in

Latin America , Princeton: Princeton University Press.

Ceku, Ethem. 2003. Mendimi Politik i Levizjes llegale ne Kosove 1945-198/,

Prishting: Brezi’81. Qeku, Ethem. 2004. Shekulli i Ilegales: Proceset gjuqesore kunder ilegales ne

Kosove (Dokumente), Prishtine: Brezi’81.

Daskalovski, £idas. 2005. Walking on the Edge: Consolidating Multiethnic

Macedonia 1989-2004, Skopje: Dominant. Davenport, Christian, ed. 2000. Paths to State Repression: Human Rights

Violations and Contentious Politics, Lanham: Rowman &

Littlefield.

Davis, Ian. 2002. Small Arms and Light Weapons in the Federal Republic o f

Yugoslavia: The Nature o f the Problem, London: Saferworld.

Dawisha, Karen & Bruce Parrott, eds. 1997. Politics, Power, and the Struggle

for Democracy in South-East Europe, Cambridge: Cambridge

University Press.

De Vos, George & Lola Romanucci-Ross. eds. 1982. Ethnic Identity: Cultural

Continuities and Change, Chicago: University of Chicago Press.

Dekanid, Igor. 2004. Demokratizacija Hrvatske: Uspjesi i Kompleksi

Suvremene Hrvatske Politike, Zagreb: Prometej.

268 Despres, Leo. ed 1975. Ethnicity and Resource Competition in Plural Societies,

Hague: Mouton.

Divjak, Jovan. 2001. “The First Phase, 1992-1993: struggle for survival and

genesis of the Army of the Bosnia-Herzegovina,” in Maga§ & Zanic

2001: 152-177.

Dobkowski, Michael N. & Isidor Wallimann. 1992. Genocide in Our Time: an

Annotated Bibliography with Analytical Introductions, Ann Arbor:

Pierian.

Dudley, Riyan & Ross A. Miller. 1998. “Group Rebellion in the 1980s,” The

Journal o f Conflict Resolution, 42-1: 77-96.

Eckstein, Harry. 1965. “On the Etiology of Internal Wars,” History and Theory, 4-2: 133-163.

Fearon, James D. 1991. “Counterfactuals and Hypothesis Testing in Political

Science,” World Politics, 43-2: 169-195.

Fearon, James D. 1998. “Commitment Problems and the Spread of Ethnic Conflict,” in Lake & Rothchild 1998: 107-126.

Fearon, James D. 2004. “Separatist Wars, Partition, and World Order,” Security

Studies, 13-4: 394-415.

Fearon, James D. & David D. Laitin. 1999. “Weak States, Rough Terrain, and

Large-Scale Ethnic Violence since 1945,” Paper presented at the

Annual Meetings of the American Political Science Association,

http://www. stanford.edu/group/ethnic/workingpapers/insurgl .pdf

Fearon, James D. & David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil

War,” American Political Science Review, 97-1: 75-90.

Felson, Richard B. & James T. Tedeschi. eds. 1993. Aggression and Violence:

Social Interactionst Perspectives, Washington, D.C.: American

Psychological Association.

269 Gagnon, Valere P. Jr. 1994. “Ethnic Nationalism and International Conflict:

The Case of Serbia,” International Security, 19-3: 130-166.

Gagnon, Valere P. Jr. 2004. The Myth o f Ethnic War: Serbia and Croatia in the

1990s, Ithaca: Cornell University Press.

Gagnon, Valere P. Jr. 2007. “Response: February 19, 2007,” Southeastern

Europe, Vol. 31-32, pp. 217-224.

Geddes, Barbara. 1990. “How the Cases You Choose Affect the Answers You

Get,” Political Analysis, 2: 131 -150.

Geddes, Barbara. 2003. Paradigms and Sand Castles: Theory Building and

Research Design in Comparative Politics, Ann Arbor: University of

Michigan Press. Glazer, Nathan & Daniel P. Moynihan. eds. 1975. Ethnicity: Theory and

Experience, Cambridge: Harvard University Press.

Glenny, Misha. 1996. The Fall o f Yugoslavia, 3rd edition, London: Penguin

Books. Glenny, Misha. 2002. “Elections in Macedonia: not many dead,” http://www.

opendemocracv.net/debates/article-2-42-380.isp

Goati, Vladimir. 1992. “Politicki Zivot Bosne i Hercegovine 1989-1992,” in

Janjic & Shoup 1992: 48-61.

Goati, Vladimir. 2001. Izbori u SRJ od 1990. do 1998.: Volja Gradana ili

Izborna Manipulacija, II dopunjeno izdanje, Beograd: Centar za

Slobodne Izbore i Demokratiju.

Goldthorpe, John H. 2001. “Causation, Statistics, and Sociology” European

Sociological Review, 17-1: 1-20.

Gow, James. 2003. The Serbian Project and Its Adversaries: A Strategy o f War

Crimes, London: Hurst.

Grdesic, Ivan, Miijana Kasapovi6, Ivan Siber & Nenad ZakoSek. 1991.

270 Hrvatska u izborima VO., Zagreb: Naprijed.

Grimshaw, Allen D. 1969. Racial Violence in the United States, Chicago:

Aldine.

Gurr, Ted Robert. 1970. Why Men Rebel, Princeton: Princeton University

Press.

Gurr, Ted Robert. 1993. Minorities at Risk: A Global View o f Ethnopolitical

Conflicts, Washington, D.C.: U.S. Institute of Peace.

Gurr, Ted Robert. 2000. Peoples versus States: Minorities at Risk in the New

Century, Washington, D.C.: U.S. Institute of Peace.

Hadzic, Miroslav, ed. 2004. The Violent Dissolution o f Yugoslavia: Causes,

Dynamics and Effects, Beograd: Centre for Civil-Military Relations. Hall, Raymond L. ed. 1979. Ethnic Autonomy-Comparative Dynamics: The

Americas, Europe and the Developing World, New York: Pergamon.

Hamzaj, Bardh. 2000. A Narrative About War and Freedom (Dialog with the

commander Ramush Haradinaj), Prishtina: Zeri. HCHRS (Helsinki Committee for Human Rights in Serbia). 1997. Report on

Human Rights in Serbia for 1996, Beograd: Prometej.

HCHRS. 1998. Kosovo: pravo i politika, Kosovo u Normativnim Aktima Pre i

Posle 1974. godine, Beograd: Zagorac.

Heiberg, Marianne, Brendan O’Leary & John Tirman. eds. 2007. Terror,

Insurgency, and the State: Ending Protracted Conflicts,

Philadelphia: University of Pennsylvania Press.

Heraclides, Alexis. 1991. The Self-Determination of Minorities in International

Politics, London: Frank Cass.

Hibbs, Douglas A. 1973. Mass Political Violence: A Cross-national Causal

Analysis, New York: Wiley-Interscience.

Hislope, Robert. 1996. “Intra-Ethnic Conflict in Croatia and Serbia: Flanking

271 and the Consequences for Democracy,” East European Quarterly,

30-4: 471-494.

Hoare, Marko Attila. 2004. How Bosnia Armed, London: Saqi Books.

Hockenos, Paul. 2003. Homeland Calling: Exile Patriotism and the Balkan

Wars, Ithaca: Cornell University Press.

Horowitz, Donald L. 1985. Ethnic Groups in Conflict, Berkeley: University of

California Press.

Horowitz, Donald L. 2001. The Deadly Ethnic Riot, Berkeley: University of

California Press.

HRW (Human Rights Watch). 2001. Macedonia - Crimes Against Civilians:

Abuses by Macedonian Forces in Ljuboten, August 10-12, 2001,

http://www. hrw. org/reports/2001/macedoni a/

Humphreys, Macartan. 2005. “Natural Resource, Conflict, and Conflict

Resolution: Uncovering the Mechanisms” Journal o f Conflict

Resolution, 49-4: 508-537. Ibrahimagid, Omer. 2001. Srpsko Osporavanje Bosne i Bosnjaka, Sarajevo:

Magistrat.

ICG (International Crisis Group). 1998a. Kosovo Spring, ICG Report,

http://www. crisis group, org/

ICG 1998b. The View from Tirana: The Albanian Dimension of the Kosovo

Crisis, ICG Balkans Report No. 36.

ICG 1998c. Kosovo's Long Hot Summer: Briefing on military, humanitarian

and political developments in Kosovo, ICG Balkans Report No. 41.

ICG. 1999. Who’s Who in Kosovo, ICG Balkans Report No. 76.

ICG 2000. Macedonia's Ethnic Albanians: Bridging the Gulf, ICG Balkans

Report No. 098.

ICG 2001. The Macedonian Question: Reform or Rebellion, ICG Balkans

272 Report No. 109.

ICG. 2004. The Collapse in Kosovo, ICG Europe Report No. 155.

IICK (Independent International Commission on Kosovo). 2000. Kosovo

Report: Conflict, International Response, Lessons Learned, Oxford:

Oxford University Press.

IWPR (Institute of War and Peace Reporting). 2004. The 2001 Conflict in

FYROM: Reflections, London: Conflict Studies Research Centre,

http://www.da.mod.uk/CSRC/documents/balkans/

Janev, Goran. 2003. “Kosovo Independence and Macedonian Stability: Is

There Any Alternative to the Nationalistic Discourse?” in Bieber &

Daskalovski 2003: 302-320. Janie, Dusan & Paul Shoup, eds. 1992. Bosna i Hercegovina Izmedju Rata i

Mira, Beograd: Dom Omladine.

Jenkins, J. Craig. 1983. “Resource Mobilization Theory and the Study of Social Movements” Annual Review o f Sociology, 9: 527-553.

Jovic, Dejan. 2001. “Fear of Becoming Minority as a Motivator of Conflict in

the Former Yugoslavia,” Balkanologie, 5-1 &2:21-36.

Judah, Tim. 2000a. Kosovo: War and Revenge, New Haven: Yale University

Press. Judah, Tim. 2000b. The Serbs: History, Myth, and the Destruction o f

Yugoslavia, 2nd edition, New Haven: Yale University Press.

Judah, Tim. 2001. “The Growing Pains of the Kosovo Liberation Army,” in

Waller etal 2001:20-24.

Kadijevic, Veljko. 1993. Moje Videnje Raspada: Vojska Bez Drzave, Beograd:

Politika.

Kaldor, Mary. 2001. New and Old Wars: Organised Violence in a Global Era,

Stanford: Stanford University Press.

273 Kaleb, Jadranko. 1999. Zadar u Domovinskom Ratu 1990.-1991., Zadar: Rujan.

Karanfilova-Panovska, Fani & Roska Ilijevska. eds. 2005. Inventory

Macedonia. December 1999-December 2001, Skopje: Foundation

Open Society Institute - Macedonia.

Katznelson, Ira & Helen V. Milner, eds. 2002. Political Science: The State of

the Discipline , 3rd edition, New York: W. W. Norton.

King, John C. 2000. “Exploring the Ameliorating Effects of Democracy on

Political Repression: Cross-National Evidence,” in Davenport 2000:

217-239.

King, Gary, Robert O. Keohane & Sidney Verba. 1994. Designing Social

Inquiry: Scientific Inference in Qualitative Research , Princeton: Princeton University Press.

Kiri§ci, Kemal & Gareth M. Winrow. 1997. The Kurdish Question and Turkey:

An Example of a Trans-state Ethnic Conflict, London: Frank Cass.

Klier, John D. & Shlomo Lambroza. eds. 1992. Pogroms: Anti-Jewish Violence

in Modem Russian History, Cambridge: Cambridge University

Press.

Kola, Paulin. 2003. The Search for Greater Albania, London: Hurst.

Kostovidova, Denisa. 1996. Parallel Worlds: Response o f Kosovo Albanians to

the Loss o f Autonomy in Serbia 1989-1996, Keele: Keele European

Research Centre.

Kresovic, Tomislav, Jelena Obradovic-Savuljesko, Miodrag Dinic & Gorica

Dokic. 1998. Kosovo Dossier: Through Terrorism to Independence,

Beograd: BINA.

Kubo, Keiichi. 2004. “Democratization and Inter-ethnic Relations in

Multiethnic Countries: A Comparative Analysis of Croatia and

Macedonia,” Acta Slavica Iaponica, 21: 181 -201.

274 Kubo, Keiichi. 2007. “The Radicalisation and Ethnicization of Elections: The

1990 Local Elections and the Ethnic Conflict in Croatia,”

Ethnopolitics, 6-1:21-41.

Laitin, David D. 1995. “Transitions to Democracy and Territorial Integrity,” in

Przeworski 1995: 19-33.

Laitin, David D. 1998. Identity in Formation: The Russian-Speaking

Populations in the Near Abroad, Ithaca: Cornell University Press.

Laitin, David D. 2001a. “Secessionist Rebellion in the Former Soviet Union,”

Comparative Political Studies, 34-8: 839-861.

Laitin, David D. 2001b. “Book Review: Donald L. Horowitz, The Deadly

Ethnic Riot. Berkeley: University of California Press, 2001.” Comparative Political Studies, 34-9: 1092-1099.

Laitin, David D. 2002. “Comparative Politics: The State of the Subdiscipline,”

in Katznelson & Milner 2002: 630-659.

Lake, David D. & Donald Rothchild. eds. 1998. The International Spread of Ethnic Conflict: Fear, Diffusion and Escalation, Princeton:

Princeton University Press.

Lampe, John R. 1996. Yugoslavia as History: Twice There Was a Country,

Cambridge: Cambridge University Press.

Letica, Slaven & Mario Nobilo. 1991. JNA: Rat Protiv Hrvatske, Zagreb:

Globus.

Lichbach, Mark Irving. 1998. The Rebel's Dilemma, Ann Arbor: University of

Michigan Press.

Lieberman, Evan. 2005. “Nested Analysis as a Mixed-Method Strategy for

Comparative Research,” American Political Science Review, 99-3:

435-452.

Lijphart, Arend. 1971. “Comparative Politics and the Comparative Method,”

275 American Political Science Review, 65-3: 682-693.

Long, J. Scott. 1997. Regression Models for Categorical and Limited

Dependent Variables, London: Sage.

Long, J. Scott & Jeremy Freese. 2003. Regression Models for Categorical

Dependent Variables Using Stata, revised edition, Lakeway Drive:

Stata Press.

Lukic, Dejan & Pero Simic. 2001. Vatre iPotop, Beograd: Filip ViSnjic.

Magas, Branka & Ivo Zanic. eds. 2001. The War in Croatia and

Bosnia-Herzegovina 1991-1995, London: Frank Cass.

Marijan, Davor. 2002. Smrt Oklopne Brigade: Prilozi za Istrazivanje Rata za

Hrvatsku i Bosnu i Hercegovinu 1990.-1992., Zagreb: Naklada

ZORO. Marijan, Davor. 2004. Bitka za Vukovar, Zagreb: Hrvatski Institut za Povijest.

Marko, Joseph & Tomislav Boric, eds. 1991. Slowenien-Kroatien-Serbien: Die

Neuen Verfassungen, Wien: Bohlau.

McCarthy, John D. & Mayer N. Zald. 1977. “Resource Mobilization and Social

Movements: A Partial Theory” American Journal of Sociology,

82-6: 1212-1241.

Metz, Daniel. 2005. “Methodology: Case Selection and Variable Coding,”

httn:// web.mit. edu/people/dmetz/method. doc

Mihailovic, Srecko. ed. 1997. Izmedu Osporavanja i Podrske: Javno Mnenje o

Legitimitetu Trece Jugoslavi/e, Beograd: Institut druStvenih nauka.

Mihailovic, Sredko. 1999. “Virtuelna Legitimacija Trece Jugoslavije,” in

Slavujevic & Mihailovic 1999: 137-252.

Mijalkovski, Milan & Petar Damjanov. 2002. Terorizam Albanskih Esktremista,

Beograd: Vojska.

Miko, Jason. 2004. Boris Trajkovski: President o f the Republic o f Macedonia,

276 1999-2004, Skopje: The Boris Trajkovski International Foundation.

Milosavlevski, Slavko & Mirche Tomovski, 1997. Albanians in the Republic of

Macedonia 1945-1995: Legislative, Political documentation,

Statistics, Skopje: NIP “Studentski Zbor”.

Milosevic, Slobodan. 1989. Godine Raspleta, trece izdanje, Beograd:

Beogradski Izdava£ko-Grafi£ki Zavod.

Nilsen, Kristijan A. 2004. “Ministarstvo unutraSnjih poslova bosanskih Srba:

Geneza, ucinak, komanda i kontrola, 1990-1992,” in Biserko 2004: 281-392.

Norris, John. 2005. Collision Course: NATO, Russia, and Kosovo, Westport: Praeger.

Oberg, Magnus. 2002. The Onset o f Ethnic War as a Bargaining Process:

Testing a Costly Signaling Model, Uppsala: Department of Peace

and Conflict Research, Uppsala University.

Oberschall, Anthony. 1973. Social Conflict and Social Movements, Englewood

Cliffs: Prentice-Hall.

O’Leary, Brendan & John Tirman, “Thinking About Durable Political

Violence,” in Heiberg et al 2007: 1-17.

Ordanoski, Saso. 2004. “Lions and Tigers: The Militarisation of the

Macedonian Right,” in IWPR 2004: 17-28.

Owen, David. 1995. Balkan Odyssey, San Diego: Harcourt Brace.

Perry, Duncan M. 1997. “The Republic of Macedonia: finding its way,” in

Dawisha & Parrott 1997:226-281.

Petersen, Roger D. 2001. Resistance and Rebellion: Lessons from Eastern

Europe, Cambridge: Cambridge University Press.

Petersen, Roger D. 2002. Understanding Ethnic Violence: Fear, Hatred, and

Resentment in Twentieth-century Eastern Europe, Cambridge:

277 Cambridge University Press.

Petrid, Volfgang & Robert Pihler. 2002. Dugi Put u Rat: Kosovo i

Medunarodna Zajednica, 1989-1999., Beograd: Samizdat B92.

Pettifer, James. 2001. “Former Yugoslav Macedonia: The Shades Of Night?”

CSRC Research Paper, http://www.da.mod.uk/CSRC/

Pettifer, James. 2004. “Introduction,” in IWPR 2004.

Phillips, John. 2004. Macedonia: Warlords and Rebels in the Balkans, London:

I.B. Tauris.

Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of

Politics,” American Political Science Review, 94-2: 251-267.

Pipa, Arshi. 1989. “The Political Situation of the Albanins in Yugoslavia, with Particular Attention to the Kosovo Problem: A Critical Approach,”

East Eurpean Quarterly, 23-2: 159-181.

PleStina, Dijana. 1992. Regional Development in Communist Yugoslavia:

Success, Failure, and Consequences, Boulder: Westview. Pojar, Daniel J. Jr. 2005. Lessons Not Learned: The Rekindling o f Thailand’s

Pattani Problem, PhD Thesis, Naval Postgraduate School, USA,

http://www.ccc.nps.naw.mil/

Popetrevski, Vasko & Veton Latifi. 2004. “The Ohrid Framework Agreement

Negotiations,” in IWPR 2004: 29-36.

Posen, Barry R. 1993. “The security dilemma and ethnic conflict” in Brown

1993: 103-124.

Poulton, Hugh. 2000. Who are the Macedonians? 2nd edition, London: Hurst.

Przeworski, Adem. ed. 1995. Sustainable Democracy, Cambridge: Cambridge

University Press.

Przeworski, Adam & Henry Teune. 1970. The Logic o f Comparative Social

Inquiry, New York: Wiley.

278 Qosja, Rexhep. 1990. Nezasticena Sudbina: O Albanicima u Jugoslaviji danas,

Zagreb: Hrvatska Socijalno Liberalna Stranka.

Rabushka, Alvin & Kenneth A. Shepsle. 1972. Politics in Plural Societies: a

Theory o f Democratic Instability, Columbus: Charles E. Merrill.

Ramet, Sabrina R & Vjeran Pavlakovic, eds. 2005. Serbia Since 1989: Politics

and Society under Milosevic and After, Seattle: University of

Washington Press. Rimac, Ivan, Stanko Rihtar & Maria Oliveira-Roca. 1992. “Multivarijatna

klasifikacija opcina Hrvatske kao moguca metoda regionalizacije

Republike,” Drustvena Istrazivanja, 1-1: 87-99.

Ron, James. 2003. Frontiers and Ghettos: State Violence in Serbia and Israel,

Berkeley: University of California Press. Roudometof, Victor. 2002. Collective Memory, National Identity, and Ethnic

Conflict: Greece, Bulgaria, and the Macedonian Question,

Westport: Praeger. RS (Republika Srbija). 2003. Albanski Terorizam i Organizovani Kriminal na

Kosovu i Metohiji, Beograd: Vlada Republike Srbije.

Rusi, Iso. 2004. “From Army to Party: The Politics of the NLA,” in IWPR

2004: 1-16.

Sambanis, Nicholas. 2001. “Do Ethnic and Nonethnic Civil Wars Have the

Same Causes? - A Theoretical and Empirical Inquiry (Part I),”

Journal o f Conflict Resolution, 45-3: 259-282.

Schmidt, Vivien A. 2002. “Europeanization and the mechanics of economic

policy adjustment,” Journal o f European Public Policy, 9-6:

894-912.

Sekelj, Laslo. 2000. “Parties and Elections: The Federal Republic of

Yugoslavia. Change without Transformation,” Europe-Asia Studies,

279 52-1: 57-75.

Sell, Louis. 2002. Slobodan Milosevic and the Destruction o f Yugoslavia,

Durham: Duke University Press.

Silber, Laura & Allan Little. 1996. The Death o f Yugoslavia, revised edition,

London: Penguin Books.

Simic, Predrag. 2000. Put u Rambuje: Kosovska Kriza 1995-2000., Beograd:

Nea.

Simons, Anna & John Mueller. 2001. “Correspondence: The Dynamics of

Internal Conflict,” International Security, 25-4: 187-192.

Sisk, Timothy D. 1996. Power Sharing and Internaitonal Mediation in Ethnic

Conflicts, Washington D.C.: U.S. Institute of Peace. Slavujevic, Zoran D. 1997. “Kriza Poverenja u Institucije Sistema,” in MihailoviC 1997: 64-94.

Slavujevic, Zoran D. 1999. “Delegitimizacija Sistema i Njegovih Institucija,”

in Slavujevic & Mihailovic 1999: 9-136. Slavujevic, Zoran D. & Srecko Mihailovic. 1999. Dva Ogleda o Legitimitetu:

Javno Mnenje o Legitimitetu Trece Jugoslavije, Beograd: Institut

druStvenih nauka.

Smith, Anthony D. 1986. The Ethnic Origins o f Nations, Oxford: B. Blackwell.

Soltan, Karol, Eric M. Uslaner & Virginioa Haufler. eds. 1998. Institutions and

Social Order, Ann Arbor: University of Michigan Press.

Stack, John F. Jr. ed. 1986. The Primordial Chalenge: Ethnicity in the

Contemporary World, Westport: Greenwood Press.

Sterling, Richard W. 1979. “Ethnic Separatism in the International System,” in

Hall 1979: 413-425.

Sullivan, Stacy. 2004. Be not afraid, for you have sons in America: how a

Brooklyn roofer helped lure the U.S. into the Kosovo New war,

280 York: St. Martin’s.

Simac, Petar. 2001. Svjedocanstva o Domovinskom Ratu u Dalmaciji 1991.,

Split: Franjo Kluz.

Spegelj, Martin. 2001a. “The First Phase, 1990-1992: the JNA prepares for

aggression and Croatia for defence,” in Magas & Zanic 2001: 14-40.

Spegelj, Martin. 2001b. Sjecanja Vojnika, Zagreb: Znanje.

Sterc, Stjepan. 1992. “Prostomi i demografski aspekti revitalizacije ruralnih

naselja u Hrvatskoj, ” Drustvena Istrazivanja, 1-1: 127-157.

Tahiri, Bedri. 2006. Adem Jashari: Legjende e Legjendave, Prishtine: Rilindja.

Tanner, Marcus. 2001. Croatia: A Nation Forged in War, 2nd edition, New

Haven: Yale University Press. Tedeschi, James T. & Mitchell S. Nesler. 1993. “Grievances: Development and

Reactions,” in Felson & Tedeschi 1993: 13-45.

Tha?i, Toger Naser. 2004. Memorial Complex: Adem Jashari, Prekaz,

Prishtine: Shtypshkronja Dukagjini.

Thomas, Robert. 1999. The Politics in Serbia in the 1990s, New York:

Columbia University Press.

Trajkovski, Boris. 2005. “The Framework Agreement - Solution to the Crisis

and Step Forward Towards Europe,” in Karanfilova-Panovska &

Ilijevska2005: 53-56.

Trix, Frances. 2005. “Kosovar Albanians between a Rock and a Hard Place,” in

Ramet & Pavlakovi6 2005: 309-349.

Troebst, Stefan. 1998. Conflict in Kosovo: Failure o f Prevention? An

Analytical Documentation, 1992-1998, Flensburg: European Centre

for Minority Issues.

Tus, Anton. 2001. “The War in Slovenia and Croatia up to the Sarajevo

Ceasefire,” in Maga§ & Zanic 2001: 41-66.

281 Van Den Berghe, Pierre L. 1990. State Violence and Ethnicity , Niwot:

University of Colorado Press.

Van Evera, Stephen. 1997. Guide to Methods for Students o f Political Science,

Ithaca: Cornell University Press.

Varshney, Ashutosh. 2001. “Ethnic Conflict and Civil Society: India and

Beyond,” World Politics, 53: 362-398.

Varshney, Ashutosh. 2002.Ethnic conflict and Civic Life: Hindus and Muslims

in India , New Haven: Yale University Press.

Vickers, Miranda. 1998. Between Serb and Albanian: a History o f Kosovo,

London: Hurst.

Waller, Michael, Kyril Drezov & Biilent Gokay, eds. 2001. Kosovo: The Politics o f Delusion, London: Frank Cass.

Weingast, Barry R. 1998. “Constructing Trust: The Political and Economic

Roots of Ethnic and Regional Conflicts,” in Soltan et al 1998:

163-200. Wilson, Alfred Jeyaratnam. 1988. The Break-Up o f Sri Lanka: The

Sinhalese-Tamil Conflict, London: Hurst.

Wolff, Stefan. 2003. “The Limits of Non-Military International Intervention: A

Case Study of the Kosovo Conflict,” in Bieber & Daskalovski 2003:

79-100.

Woodward, Susan L. 1995. Balkan Tragedy: Chaos and Dissolution After the

Cold War, Washington, DC: The Brookings Institution.

Yannis, Alexandras. 2003. “Kosovo: The Political Economy of Conflict and

Peacebuilding,” in Ballentine & Sherman 2003: 167-195.

Zimmermann, Warren. 1996. Origins of a Catastrophe: Yugoslavia and Its

Destroyers, New York: Times Books.

282 Appendices

Appendix 1: Scale for the Anti-regime rebellion variable (REBEL) of the MAR dataset

0 - None reported 1 - Political banditry, sporadic terrorism 2 - Campaigns of terrorism 3 - Local rebellions: Armed attempts to seize power in a locale 4 -Small-scale guerrilla activity: All of the following must exist: 1) fewer than 1000 armed fighters; 2) sporadic armed attacks (less than six reported per year); and 3) attacks in a small part of the area occupied by the group, or in one or two other locales. 5 - Intermediate guerrilla activity: Has one or two of the defining traits of large-scale activity and one or two of the defining traits of small-scale activity 6 - Large-scale guerrilla activity: All of the following must exist: 1) more than 1000 armed fighters; 2) frequent armed attacks (more than 6 per year); and 3) attacks affecting a large part of the area occupied by the group 7 - Protracted civil war: Fought by rebel military units with base areas. Missing - No basis for judgment

283 Appendix 2: Scale for the military repression variables (REP20, REP21, Rep22) of the MAR dataset

0 - No repression reported 1 - Tactics used against group members who are engaged in collective action participating in marches, preparing for or carrying out armed attacks, etc.).

2 - Tactics used against group members in both kinds o f circumstances engaged and not engaged in collective action), or in ambiguous situations, for example when it is not clear from source materials whether repressive action was provoked or not.

3 - Tactics used against group members who are not engaged in collective action.

Appendix 3: Ordered Logit Analysis of the Determinants of Military Repressions, 1996-2000 (Dependent Variable: REP21 and REP22)

model 3 model 4 Dependent Variable RE P21 RE P21 Independent Variables coef. st d. err coef. std err Repression of the Previous Year 1.81*** 0.62 1 85*** 0.57 (LDV) Rebellion of the Previous Year 0 062 0 43 Low-Intensity Rebellion in the -0.15 1.35 Previous Year High-Inteiisity Rebellion in the 417*** 0.77 406*** 0.7 Previous Year Democracy -0.028 0046 -0025 0.042 Level of Economic Development 0068 0 056 0061 0053 Cut Point (1) 5 9 0.64 5 8 061 Cut Point (2) 7 38 1.1 7 29 1 13 Cut Point (3) 8.23 1.36 8 15 1.41 N 967 967

284 model 5 model 6 Dependent Variable REP22 r e : Independent Variables coef. std.en. coef. std.en. Repression of the Previous Year 1 96*** 0.33 1.96*** 0.33 (LDV) j 0 Rebellion of the Previous Year 0.43* 1 Low-Intensity Rebellion in the 1.86* 0.97 Previous Year High-Intensity Rebellion in the 3.34*** 0.95 3.62*** 1.25 Previous Year Democracy -0 12** 0.059 -0.09 0 061 Level of Economic Development -0.09" 0.14 -0.2 0.2 Cut Point (1) 5 8? 0.73 5.93 0.9 Cut Point (2) 5 96 0.73 VO o i 0.93 r r i 00

Cut Point (3) 7.19 11 1.28 N 968 968 * p < 0. k ** p < 0.05; *** p < 0.01

Appendix 4: Random-Effects Logit Analysis of Determinants of the Escalation of Rebellion, 1996-2000

Model 3 Model 4 coef. std.err. coef std.en. Military Repression (REP21) -0 59 0.54 Military Repression (REP22) 0.15 0.21 Political Discrimination 0.0~8 0.13 0.071 0.13 Economic Discrimination 0.26* 0 1 4 0.23 0.15 Lost Autonomy -0 2 0.15 -0.19 0.15 Duration of Peaceful Protest -0 007 0 011 -0 008 0 011 Dictatorship -0 14 0 33 -0 19 0.33 Mountainous Terrain -0 014 0 11 -0.017 0 11 Political Instability 0.07 0 34 0.08 0.34 Level of Economic Development -0.21*** 0.07 -021*** 007 Regional Concentration -0.69 042 -0 58 0.43 -4

Constant -1.17** P -1 26** 0.57 N 659 660 * p <0.1; ** p < 0.05; ***p<0.01

285 Appendix 5: Ordered Logit Analysis of the Determinants of Level of Rebellion, 1996-2000 (Dependent Variable: REB)

Model 3 Mod el 4 Independent Variables coef std err coef. std err Level of Rebellion in the Previous 1 1 -»* * * 0.11 113*** 0.12 Year (LDV)

Military Repression (REP21) 0 54** 026 Military Repression (REP22) 020 021 Politleal Disci miina t ion 0 10 0 11 0.10 011 Economic Discrimma t ion 0.0^ 0.11 0 06 0.11 Lost Autonomy -o .o s 0.09 -0 09 0.09 Duration ofPeacefi.il Protest 0 014*** 0.0053 0 011** 0005 Dictatorsliip 0 13 0 26 007 027 Moimtamous Terrain -002 007 -0 03 007 Political Instability 0 16 0 22 0 06 0 22 Level of Economic Development -0 04 003 -0 03 003 Regional Concentration 0 11 0 33 0 16 0.34 Cut Point (1) 2 08 042 2 03 0 42 Cut Point (2) 3.07 0 43 3 03 042 Cut Point (3) 3.66 049 363 048 Cut Point (4) 467 054 464 0 54 Cut Point (5) 5.93 061 589 0.6 Cut Point (6) 698 0 72 691 0 71 Cut Point (7) 8 32 084 8 17 083 N 658 659 * p 01; ** p < 0.05; *** p < 0 01

Appendix 6: Number of Votes cast for Albanian candidates in the first round and the increase of votes received by two final candidates in the second round, 1999 Presidential Elections in Macedonia

Number of Votes cast for Two Increase of the Votes m the Secound Electoral Albanian Candidates in the Round, received by; Unit First Round Trajkovski Petkovski 1 18 2104 1181 2 19 726 1137 3 14 2309 1420 4 15 2027 1953

286 5 7 2170 1586

6 15 2 1 1 1 2259

7 2 2 2 2 1 0 1826

8 16 2051 1775 9 15 1382 2491

1 0 5 1831 1317

11 1 0 1958 1156

1 2 2 1916 1044 13 3 1855 844

14 1 0 1480 1638 15 874 1726 1406

16 1 0 728 1817 17 1498 2341 1952 18 2817 4138 1773 19 362 1285 2013

2 0 1346 2635 1754

21 15081 16313 332

2 2 2 1 1727 1653 23 13 2587 1580 24 23 1819 3131

25 1 2 2147 5003

26 6 2444 4662 27 1545 3322 2251 28 16 1680 2188

29 1 2 0 1799 2034 30 243 1804 3065 31 597 2656 2325 32 28 2230 2333 33 379 1916 2283 34 523 2433 1143 35 23 1790 2682

36 2 2 1742 2384 37 24 1932 2097 38 704 3011 1800 39 858 2703 1551

287 40 1056 2502 1447 41 1178 2806 1241 42 9522 14040 91 43 4612 9833 1241 44 12077 15612 -36 45 3193 5844 1137 46 976 1935 1443 47 461 2335 2272

48 335 2 2 1 1 2113

49 4 1 2 0 0 1402 50 942 1595 1563 51 1249 3055 1942 52 13867 16198 -67 53 9994 15126 353 54 4254 6303 858 55 10774 17890 564 56 18476 19774 283 57 2348 3434 2781 58 8336 11705 727 59 11974 18280 87 60 14716 16376 87 61 2509 6494 517 62 188 1854 3948 63 300 1893 3272 64 6111 9416 1072 65 356 2638 2071

6 6 5555 11376 1023 67 24 1410 4615

6 8 59 1437 4755 69 87 1629 3937 70 47 1410 4457

71 8 8 1741 3490

72 2 2 1016 4179

73 1 2 2 2 2293 2599 74 277 1229 3643

288 75 4185 6933 737 76 84 1149 4473 77 183 1115 4021 78 749 2098 3575 79 1179 2783 3217 80 375 1747 2888

81 11139 13367 1 1 0 1

82 2 2 2 2 3682 2560 83 1278 2719 2883 84 3483 5492 1597 85 2327 2861 1995 Source: Made by the Author, based on the election results announced by the Election Committee of Macedonia. See its website at http://www. izbori. gov.mk/

289

0 0 0 0 0 0 0 0 0 0 POLARM 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Organisation POLTCIAN Yucpa Tribe Yucpa Yanomani Indians Yanomani MISURASATA (Mascot, Rama, and Sandinista United) Sandinista and Rama, (Mascot, MISURASATA (EGTK) Katari Tupaj Guerrillero Ejercito Guatemala National Revolutionary Unit (URNG) Unit Revolutionary National Guatemala Tehuantepec (COCEI) Tehuantepec Kuna Tribesmen in Digir Community Digir in Tribesmen Kuna etc Lame, Quintin M-19, American Indian Movement Indian American (FLQ) Front Liberation Quebec Worker-Peasant-Student Coalition of the Isthmus of Isthmus the of Coalition Worker-Peasant-Student Mohawk Warriors (EZLN) Mohawk Army Liberation National Zapatista Black Panthers, etc Panthers, Black Party Nationalist Rican Puerto 1 1 1 1 1 1 6 reb 1980 3 1980 1960 3 1985 1994 1975 4 1994 4 1975 country year BOLIVIA 1985 NICARAGUA MEXICO MEXICO USA 1950

group INDIGENOUS PEOPLES INDIGENOUS AMAZONIAN INDIANS AMAZONIAN BRAZIL QUEBECOIS CANADA 1960 INDIGENOUS PEOPLESINDIGENOUS PEOPLESINDIGENOUS PANAMA PEOPLESINDIGENOUS COLOMBIA VENEZUELA HIGHLAND INDIGENOUS PPLS INDIGENOUS PEOPLES INDIGENOUS GUATEMALA INDIGENOUS PEOPLESINDIGENOUS CANADA 1990 3 AFRICAN-AMERICANS USA 1965 3 MAYANS fflSPANICS AMERICANSNATIVE USAZAPOTECS 1970 3 the rebellion Appendix Appendix 7: Initial Onset ofRebellion, 177 ethnic -groups organisation and the involvement of politicians, police and army in

290 LOWLAND INDIGENOUS the National Association of Coca Producers BOLIVIA 1985 1 0 0 PPLS (ANAPCOCA)

INDIGENOUS PEOPLES CHILE 1989 1 Mapuche All Lands Council 0 0

CATHOLICS IN N. IRELAND UK 1955 1 Irish Republican Army 0 0

SCOTS UK 1980 1 The Scottish National Liberation Army 0 0

BASQUES FRANCE 1975 1 Iparretarrek 0 0 Front de Liberation de la Bretagne / Armee BRETONS FRANCE 1965 2 0 0 Revolutionnaire Bretonne

CORSICANS FRANCE 1975 2 FLNC (Front de la Liberation Nationale de la Corse) 0 0

MUSLIM (NONCITIZENS) FRANCE 1995 1 GIA (Islamic Armed Group) 0 0

JURASSIANS SWITZERLAND 1960 1 Jura Liberation Front (FLJ) 0 0

BASQUES SPAIN 1965 2 Euzkadi Ta Azkatasuna (ETA) 0 0

CATALANS SPAIN 1970 1 Terra Lliure 0 0

SOUTH TYROLIANS ITALY 1955 1 Ein Tyrol 0 0

SARDINIANS ITALY 1965 1 Sardinian Autonomy Movement (MAS) 0 0

GREEKS ALBANIA 1994 1 MAVI (Northern Epirus Liberation Front) 0 0

ALBANIANS MACEDONIA 1996 1 Albanian Protesters 0 0

SERBS CROATIA 1992 7 Republika Srpska Krajina 1 1

KOSOVO ALBANIANS YUGOSLAVIA 1945 3 Rebels from Drenica Region 0 0

CROATS A YUGOSLAVIA 1970 1 Terrorists from Usta§a Remnants 0 0 SERBS BOSNIA 1992 7 Republika Srpska 1 1 Croatian Community of Bosanska Posavina, Herzeg CROATS BOSNIA 1992 7 1 1 Bosna, Central Bosnia

MUSLIMS BOSNIA 1995 7 Republic of Western Bosnia 1 1

MUSLIMS GREECE 1995 1 Turks of West Thrace 0 0 Turk Mukavemet Teskilati (TMT - Turkish Resistance TURKISH CYPRIOTS CYPRUS 1960 1 0 0 Organization) Gagauz Soviet Socialist Republic (declaration of GAGAUZ MOLDOVA 1991 3 1 0 independence)

SLAVS MOLDOVA 1991 4 Trans-Dniester 1 0

CHECHENS RUSSIA 1992 3 Chechen Republic (Declaration of Independence) 1 0

ESTONIANS USSR 1945 4 Metsavennad 0 0

LETTS/LATVIANS USSR 1945 6 Meza braji 0 0

LITHUANIANS USSR 1945 6 Forest Brothers 0 0 Ukrayins’ka Povstans’ka Armiya (Ukrainian Insurgent UKRAINIANS USSR 1945 6 0 0 Army)

AVARS RUSSIA 1992 1 Shamil People's Front 0 0

INGUSH RUSSIA 1993 2 Ingush Militias 0 0

RUSSIANS LITHUANIA 1991 1 Russian Activitists 0 0

ABKHAZIANS GEORGIA 1992 7 the Abkhaz Autonomous Republic 1 1 OSSETIANS (SOUTH) GEORGIA 1991 5 The South Ossetia Autonomous Region 1 1 Nagorno-Karabakh Autonomous Oblast > ARMENIANS AZERBAIJAN 1991 7 1 1 Nagorno-Karabakh Republic

TUAREG MALI 1960 6 Rebels of Adrar NTughas region 0 0 The Movement of Democratic Forces in Casamansais DIOLAS IN CASAMANCE SENEGAL 1985 1 0 0 (MDFC)

KEWRI MAURITANIA 1985 4 Forces Liberation Africaine de Mauritanie (FLAM) 0 0

HAUSA NIGER 1960 1 Sawaba Party 0 0

TUAREG NIGER 1985 3 the Front de Liberation de l'Air et 1'Azaouak (FLAA) 0 0

FULANI GUINEA 1960 1 Fulani Tribesmen 0 0

MALINKA GUINEA 1996 3 Coup attempts by dissident soldiers 0 1

MENDE SIERRA LEONE 1965 1 Coup led by army commander Lansana, etc 0 1

TEMNE SIERRA LEONE 1970 3 Coup attempt by John Bangura 0 1

ASHANTI GHANA 1965 1 Coup Attempts 0 0

EWE GHANA 1965 1 General Emmanuel Kotoka (Coup) 0 0

EWE TOGO 1993 2 the Collective of Democratic Opposition-2 (COD-2) 0 0 coup led by leaders of the demobilized veterans with the KABRE TOGO 1960 1 0 1 support of their Kabrai army allies.

WESTERNERS CAMEROON 1992 3 Social Democratic Front (SDF) 0 0

m o NIGERIA 1965 7 Independent State of Biafra 1 1 Ijaw National Congress, various Ijaw Youth movements, UAW NIGERIA 1995 1 0 0 etc

NORTHERNERS CHAD 1965 6 Front for National Liberation of Chad (FROLINAT) 0 0

SOUTHERNERS CHAD 1980 6 Commandos (codos) 0 1

LARI REP. OF CONGO 1960 3 Coup led by Alphonse Massemba-Debat 0 1

M'BOSHI REP. OF CONGO 1965 3 Coup led by Marien Ngouabi 0 1 DEM. REP. LUBA 1960 7 Etat Minier du Sud-Kasai 1 0 CONGO DEM. REP. LUND A, YEKE 1960 7 independent State of Katanga 1 0 CONGO DEM. REP. Hutu extremists and Interahamwe in north Kivu, merged HUTUS 1996 7 0 0 CONGO with militias from the Banyarwanda Hutu communities DEM. REP. Alliance of Democratic Forces for the Liberation of TUTSIS 1996 7 0 0 CONGO Congo-Zaire (ADFL)

ACHOLI UGANDA 1970 1 Forces led by Milton Obote 0 0

ANKOLE UGANDA 1980 7 National Resistance Army (NRA) 0 0 Baganda Government, led by Kabaka (Threat of BAGANDA UGANDA 1965 3 1 0 Secession) Uganda National Rescue Front (UNRF), Former Uganda KAKWA UGANDA 1980 6 0 0 National Army (FUNA) 1 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 00 0 0 00 1 0 0 0 0 0 0 0 0 00 0

Tigray People’s Liberation Front (TPLF) Front Liberation People’s Tigray (FRUD) The Eritrean Liberation Front (ELF) Front Liberation Eritrean The clan Gumuz Coup attempt by Somaliland Scouts Somaliland by attempt Coup the Western Somali Liberation Front (WSLF), the Ogadeni the (WSLF), Front Liberation Somali Western the Tutsi Gunman (assasination) Gunman Tutsi Coalition pour la Defense de la Republique (CDR) Republique la de Defense la pour Coalition Front pour la Restauration de lTJnite et de la Democratic la de et lTJnite de Restauration la pour Front National Liberation Front (ONLF) Front Liberation National National Front for the Liberation of Angola (FNLA) Angola of Liberation the for Front National Rwandan Patriotic Front (RPF) Front Patriotic Rwandan Four Gunmen from Military from Gunmen Four Rwenzururu Kingdom Rwenzururu Forces led by Milton Obote Milton by led Forces Hutu Army/Police Officers (Coup attempt) (Coup Officers Army/Police Hutu Karamojong Tribesmen Karamojong National Resistance Army (NRA) Army Resistance National 1 1 1 1 6 6 6 2 2 6 6 1991 1975 (ALF/P) [Party] Front 7Liberation Afar The 1975 1970 1965 1980 SOMALIA 1960 4 ETHIOPIA 1960 TANZANIA ETHIOPIA 1960 4 DJIBOUTI ETHIOPIA 1975 3 UGANDA 1965 RWANDA 1985 4 UGANDA BURUNDI 1965 UGANDA SOMALIS TIGREANS ETHIOPIA AFARS AFARS ETHIOPIA ISSAQ ERITREANS OROMOBAKONGO ETHIOPIA 1960 (OLF) Front 4Liberation Oromo The ANGOLA 1975 7 TUTSISTUTSIS BURUNDI 1965 NILO-SAHARANS KONJO/AMBA LANGI UGANDA 1970 LUGBARA/MADI ZANZIBARIS HUTUS HUTUS RWANDA 1994 KARAMOJONG

295 0 0 0 0 0 0 0 0 1 00 0 0 0 0 1 0 0 0 0 0 00 0 0 0 0 0 0 0

Socialist Forces Front (FFS) Front Forces Socialist of Saguiet el Hamara and Rio de Oro) de Rio and Hamara el Saguiet of Anya Anya movement Anya Anya POLIS ARIO Front (The Popular Front for the Liberation the for Front Rami Popular (The Front ARIO POLIS Unemployed Soldiers Unemployed groups vigilante Ahmed by Kung-fu led Populaires, Forces des Nationale Union Congress (PAC) Congress (IFP) Party Freedom Inkatha Afrikaner Weerstandsbeweging (AWB) Weerstandsbeweging Afrikaner African National Congress (ANC), Pan Africanist Pan (ANC), Congress National African African National Congress (ANC), Pan Africanist Pan (ANC), Congress National African Congress (PAC) Congress Coup attempt by a member of the Bemba royal house royal Bemba the of member a by attempt Coup Angola) UNITA (National Union for the Total Independence of Independence Total the for Union (National UNITA Lozi men gathered for the defence of the ruler the of defence the for gathered men Lozi Zimbabwean African People's Union (ZAPU) Union People's African Zimbabwean Front for the Liberation of Cabinda (FLEC) Cabinda of Liberation the for Front 1 1 1 1 6 2 2 7 2 4 1970 4 1960 1970 1970 4 1994 1994 3 1990 3 1960 1975 1975 SOUTH AFRICA SOUTH 1980 SOUTH AFRICA SOUTH 1960 AFRICA SOUTH MOROCCO ANGOLA NAMIBIA 1990 ANGOLA ZAMBIA ZIMBABWE 1965 4 ZAMBIA SOUTHERNERS SUDAN SAHARAWIS MOROCCO BERBERS ALGERIA 1965 EAST CAPRIVIANS EAST BERBERS XHOSA ZULUS MERINA MADAGASCAR EUROPEANS CABINDA OVTMBUNDU LOZI AFRICANS BLACK AFRICA SOUTH BEMEBE NDEBELE

296 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0

Islamic Movement, etc Movement, Islamic Fedayeen, etc Fedayeen, Movement of Independent Nasserites "Al-Murabitun" Nasserites Independent of Movement al-Da‘wa al-Islamiyya (Islamic Call) (Islamic al-Islamiyya al-Da‘wa Phalange, Lebanese Forces Lebanese Phalange, Militia led by Kamal Jumblatt Kamal by led Militia Abul Qasem Abul (SPLM/A) The Arab Political Cultural Organization (APCO) Organization Cultural Political Arab The Kurdish Worker’s Party (PKK) Party Worker’s Kurdish Azerbaijani Democratic Republic / Democratic Party of Party Democratic / Republic Democratic Azerbaijani independence) of (declaration Azerbaijan Kurdish People's Republic / Kurdistan Democratic Party Democratic /Kurdistan Republic People's Kurdish the Sudan People’s Liberation Movement/Army Liberation People’s Sudan the Retired Officers' Association, Bakhtiari tribesmen led by led tribesmen Bakhtiari Association, Officers' Retired 1 1 2 7 4 4 etc Militia, Amal 3 2 6 3 4 1965 1955 1955 1975 1960Organization 3Liberation Palestinian 1975 1955 1975 1945Party 4Democratic Kurdistan 1975workers 4 Agrarian Turkmen 1955 1975 31945 Tribesmen Baluch 1975 1945 1950 3 1985 4 ISRAEL LEBANON LEBANON IRAQ IRAQ LEBANON LEBANON LEBANON TURKEY IRAN IRAN IRAN SHI'IS SUNNIS ARABS SHI'IS PALESTINIANS JORDAN PALESTINIANS MARONITE CHRISTIANS MARONITE DRUZE TURKMEN ARABS IRAN KURDS KURDS BALUCHIS AZERBAIJANISBAKHTIARI IRAN KURDS IRAN NUBA SUDAN

297 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 00 0 0 0 0 0 00 0 0 0 0

Sikh militants led by Sant Jamail Bindranwale Jamail Sant by led militants Sikh SIMI (Students Islamic Movement of India), etc India), of Movement Islamic (Students SIMI movement Youth Guerrillas (trained by Pakistan), Pakistani troops Pakistani Pakistan), by (trained Guerrillas Youth Shi'ite pilgrims protesteres, Saudi Hezbollah Saudi protesteres, pilgrims Shi'ite Maoist Peoples' War Group (PWG) Group War Peoples' Maoist NNC (Naga National Council) National (Naga NNC Tensung Dhanglang Magar (the Voluntary Force for the for Force Voluntary (the Magar Dhanglang Tensung Jharkhand (MCCI), India of Centre Communist Maoist Demonstrators (Kwangju Uprising) (Kwangju Demonstrators Abdullojonov Faith) the of Defense Ethnic Tajik Militia led by Ahmed Shah Masoud, etc Masoud, Shah Ahmed by led Militia Tajik Ethnic East Turkistan Republic Turkistan East Islamist Groups (Sazman-e-Nasr, Sepah-e-Pasdaran, etc) Hezbollah Bahrain Sepah-e-Pasdaran, (Sazman-e-Nasr, Groups Islamist Fedayeen, etc Fedayeen, Ethnic Uzbek Militia Uzbek Ethnic Rebel Force led by former army colonel Mahmud colonel army former by led Force Abdumalik Rebel premier former and Khudoiberdiev 1 1 1 1 2 2 7 7 6 2 2 1980 1980 1987 1985 7 1998 5 INDIAINDIA 1960 1980 INDIA 1950 INDIAINDIA 1960 4 1997 CHINA 1950 INDIA 1965 4 SAUDI ARABIA SAUDI AFGHANISTAN AFGHANISTAN TAJIKISTAN SANTALS TRIBES SCHEDULED SIKHS TIBETANS TURKMENKASHMIRIS CHINA 1945 SHI’IS SPOTSTAJIKS BAHRAIN 1995 NAGAS HONAMESEMUSLIMS KOREA S. UZBEKS UZBEKS PALESTINIANS ISRAEL 1955 HAZARAS AFGHANISTAN 1980 7

298 1 0 0 0 0 0 0 0 0 1 1 00 0 0 00 0 0 0 0 0 0 0 0 0 0 000 0 0 0 0

Shan State Army State Shan Shanti Bahini Shanti Chin National Front (CNF) Front National Chin Karen National Union (KNU) Union National Karen Sindh National Party (SNP) Party National Sindh Sena Rashtriya Sangh Party Sangh Rashtriya Sena Kachin Independence Organization (KIO) Organization Independence Kachin New Mon State Party (NMSP) Party State Mon New Tripura Upajati Juba Samati (TUJS) Samati Juba Upajati Tripura guerrillas Afghan-loyal the All-Bodos Students Union (ABSU), the Bodo Security Bodo the (ABSU), Union Students All-Bodos the United Liberation Front of Assam (ULFA) Assam of Front Liberation United Mizo National Front (MNF) Front National Mizo North Arakan Muslim League Muslim Arakan North Movement) Zarakzai tribe, led by Nawab Nauroz Nawab by led tribe, Zarakzai (JSQM), Mahaz Qaumi Sindh Jeay Bhutan People's Party (BPP) Party People's Bhutan MQM (Mohajir Qaumi Movement, later Muttiheda Qaumi Muttiheda later Movement, Qaumi (Mohajir MQM Force (BSF) Force 1 1 1 6 4 6 6 2 4 1985 5 1960 1950 7 1985 5 1950 1975 1990 1950 1965 1980 4 2 0 0 0 1 BURMA 1955 BURMA BURMA INDIA INDIA 1980 INDIA 1950 INDIA 1989 5 BURMA PAKISTAN PAKISTANPAKISTAN 1965 3 PAKISTAN BANGLADESH BHUTAN 1990 3 SHANS BURMA MONS BURMA SINDHIS KACHINS KARENS ZOMIS (CHINS) ZOMIS CHITTAGONG HILL TRIBES HILL CHITTAGONG ASSAMESE ROfflNGYA (ARAKANESE) ROfflNGYA MOHAJIRS TR1PURAS LHOTSHAMPAS (PUSHTUNS)PASHTUNS PAKISTAN BODOS BALUCHIS HINDUS MIZOS

299 1 0 0 0 0 0 0 00 0 0 0 0 00 0 0 0 00 0 0 0 0 0 0 0 0 00 0 0

Group of Extremist Fijians Extremist of Group (FRETILIN) (PUSA) Association Ulama All-Aceh Bougainville Revolutionary Army Revolutionary Bougainville The Revolutionary Front for East Timor Independence Timor East for Front Revolutionary The Organization) Papua (Free Merdeka Papua Organisasi the military wing) military a group of ethnic Vietnamese fisherman Vietnamese ethnic of group a Irregular army led by General Vang Pao Vang General by led army Irregular (MNLF) Front Liberation National Moro CPB's army in Shan State, Wa troops Wa State, Shan in army CPB's Hill Tribes Hill Bajaraka Movement Bajaraka the Liberation Tigers of Tamil Eelam (LTTE) Eelam Tamil of Tigers Liberation the Rebellion Nyor Dusun Red Meo Insurgencies Meo Red New People's Army (Communist Party of the Philippines' the of Party (Communist Army People's New Mob's attack on Police on attack Mob's 1 1 1 3 2 6 6 6 4 6 1998 1988 1950 1960 3 1970 1975 1955 1955 4 1970 1975 7 2 0 0 0 2 0 0 0 1 SOUTH PAPUA N.G. PAPUA SRI LANKA SRI SRI LANKA SRI VIETNAM PHILIPPINES THAILAND 1945 3 PHILIPPINES CAMBODIA LAOS BURMA ACEHNESE INDONESIA BOUGANVILLEAN S BOUGANVILLEAN FUIANS FUI IGOROTS PAPUANS INDONESIA SRI LANKAN TAMILS LANKAN SRI MOROS TIMORESEEAST INDONESIA 1955 3 VIETNAMESE MONTAGNARDSMONTAGNARDS VIETNAM 1955 INDIAN TAMILS INDIAN HILL TRIBALS HILL HMONG MALAY-MUSLIMS TRIBES HILL NORTHERN THAILAND 1955

300